Run DCE after a LoopFlatten test to reduce spurious output [nfc]

[llvm-project.git] / clang / test / OpenMP / distribute_simd_private_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 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3

// RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5
// RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5
// RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK7
// RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK7

// RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
// RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1  -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
// RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
// RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11

// RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK13
// RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1  -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK13
// RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK15
// RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK15
// expected-no-diagnostics
#ifndef HEADER
#define HEADER

template <class T>
struct S {
  T f;
  S(T a) : f(a) {}
  S() : f() {}
  operator T() { return T(); }
  ~S() {}
};

template <typename T>
T tmain() {
  S<T> test;
  T t_var = T();
  T vec[] = {1, 2};
  S<T> s_arr[] = {1, 2};
  S<T> &var = test;
  #pragma omp target
  #pragma omp teams
#pragma omp distribute simd private(t_var, vec, s_arr, s_arr, var, var)
  for (int i = 0; i < 2; ++i) {
    vec[i] = t_var;
    s_arr[i] = var;
  }
  return T();
}

int main() {
  static int svar;
  volatile double g;
  volatile double &g1 = g;

  #ifdef LAMBDA
  [&]() {
    static float sfvar;

    #pragma omp target
    #pragma omp teams
#pragma omp distribute simd private(g, g1, svar, sfvar)
    for (int i = 0; i < 2; ++i) {
      g = 1;
      g1 = 1;
      svar = 3;
      sfvar = 4.0;
      [&]() {
        g = 2;
        g1 = 2;
        svar = 4;
        sfvar = 8.0;

      }();
    }
  }();
  return 0;
  #else
  S<float> test;
  int t_var = 0;
  int vec[] = {1, 2};
  S<float> s_arr[] = {1, 2};
  S<float> &var = test;

  #pragma omp target
  #pragma omp teams
#pragma omp distribute simd private(t_var, vec, s_arr, s_arr, var, var, svar)
  for (int i = 0; i < 2; ++i) {
    vec[i] = t_var;
    s_arr[i] = var;
  }
  int i;

  #pragma omp target
  #pragma omp teams
#pragma omp distribute simd
  for (i = 0; i < 2; ++i) {
    ;
  }
  return tmain<int>();
  #endif
}






#endif
// CHECK1-LABEL: define {{[^@]+}}@main
// CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK1-NEXT:    [[G1:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
// CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK1-NEXT:    store ptr [[G]], ptr [[G1]], align 8
// CHECK1-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68
// CHECK1-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
// 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:    [[G:%.*]] = alloca double, align 8
// CHECK1-NEXT:    [[G1:%.*]] = alloca double, align 8
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[SVAR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[SFVAR:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    store ptr [[G1]], ptr [[_TMP2]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 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:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP4:![0-9]+]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    store double 1.000000e+00, ptr [[G]], align 8, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[_TMP2]], align 8, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    store volatile double 1.000000e+00, ptr [[TMP8]], align 8, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    store i32 3, ptr [[SVAR]], align 4, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    store float 4.000000e+00, ptr [[SFVAR]], align 4, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
// CHECK1-NEXT:    store ptr [[G]], ptr [[TMP9]], align 8, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[_TMP2]], align 8, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
// CHECK1-NEXT:    store ptr [[SVAR]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 3
// CHECK1-NEXT:    store ptr [[SFVAR]], ptr [[TMP13]], align 8, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(32) [[REF_TMP]]), !llvm.access.group [[ACC_GRP4]]
// 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:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP4]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK1-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[G1:%.*]] = alloca ptr, align 4
// CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
// CHECK3-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK3-NEXT:    store ptr [[G]], ptr [[G1]], align 4
// CHECK3-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK3-NEXT:    ret i32 0
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68
// CHECK3-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined)
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined
// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP1:%.*]] = alloca ptr, 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 ptr, align 4
// CHECK3-NEXT:    [[SVAR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SFVAR:%.*]] = alloca float, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 4
// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    store ptr [[G1]], ptr [[_TMP2]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 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:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP5:![0-9]+]]
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    store double 1.000000e+00, ptr [[G]], align 8, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[_TMP2]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    store volatile double 1.000000e+00, ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    store i32 3, ptr [[SVAR]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    store float 4.000000e+00, ptr [[SFVAR]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
// CHECK3-NEXT:    store ptr [[G]], ptr [[TMP9]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[_TMP2]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
// CHECK3-NEXT:    store ptr [[SVAR]], ptr [[TMP12]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 3
// CHECK3-NEXT:    store ptr [[SFVAR]], ptr [[TMP13]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 4 dereferenceable(16) [[REF_TMP]]), !llvm.access.group [[ACC_GRP5]]
// 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, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP5]]
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK3-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@main
// CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK5-NEXT:    [[G1:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
// CHECK5-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK5-NEXT:    store ptr [[G]], ptr [[G1]], align 8
// CHECK5-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK5-NEXT:    ret i32 0
//
//
// CHECK7-LABEL: define {{[^@]+}}@main
// CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK7-NEXT:    [[G1:%.*]] = alloca ptr, align 4
// CHECK7-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
// CHECK7-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK7-NEXT:    store ptr [[G]], ptr [[G1]], align 4
// CHECK7-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK7-NEXT:    ret i32 0
//
//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK9-NEXT:    [[G1:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK9-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK9-NEXT:    [[VAR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[KERNEL_ARGS3:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    store ptr [[G]], ptr [[G1]], align 8
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK9-NEXT:    store i32 0, ptr [[T_VAR]], align 4
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const.main.vec, i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 0
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float noundef 1.000000e+00)
// CHECK9-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYINIT_BEGIN]], i64 1
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float noundef 2.000000e+00)
// CHECK9-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 8
// CHECK9-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 2, ptr [[TMP0]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 2, ptr [[TMP8]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP11]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK9-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93.region_id, ptr [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK9:       omp_offload.failed:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93() #[[ATTR4:[0-9]+]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4
// CHECK9-NEXT:    store i32 [[TMP15]], ptr [[I_CASTED]], align 4
// CHECK9-NEXT:    [[TMP16:%.*]] = load i64, ptr [[I_CASTED]], align 8
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP16]], ptr [[TMP17]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP16]], ptr [[TMP18]], align 8
// CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK9-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 0
// CHECK9-NEXT:    store i32 2, ptr [[TMP22]], align 4
// CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 1
// CHECK9-NEXT:    store i32 1, ptr [[TMP23]], align 4
// CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[TMP20]], ptr [[TMP24]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[TMP21]], ptr [[TMP25]], align 8
// CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 4
// CHECK9-NEXT:    store ptr @.offload_sizes, ptr [[TMP26]], align 8
// CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 5
// CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP27]], align 8
// CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP28]], align 8
// CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP29]], align 8
// CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 8
// CHECK9-NEXT:    store i64 2, ptr [[TMP30]], align 8
// CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP31]], align 8
// CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP32]], align 4
// CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP33]], align 4
// CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP34]], align 4
// CHECK9-NEXT:    [[TMP35:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 1, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.region_id, ptr [[KERNEL_ARGS3]])
// CHECK9-NEXT:    [[TMP36:%.*]] = icmp ne i32 [[TMP35]], 0
// CHECK9-NEXT:    br i1 [[TMP36]], label [[OMP_OFFLOAD_FAILED4:%.*]], label [[OMP_OFFLOAD_CONT5:%.*]]
// CHECK9:       omp_offload.failed4:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102(i64 [[TMP16]]) #[[ATTR4]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT5]]
// CHECK9:       omp_offload.cont5:
// CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v()
// CHECK9-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP37]], [[OMP_OFFLOAD_CONT5]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE6:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done6:
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK9-NEXT:    [[TMP38:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    ret i32 [[TMP38]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93
// CHECK9-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93.omp_outlined)
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93.omp_outlined
// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK9-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK9-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[SVAR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK9:       arrayctor.loop:
// CHECK9-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK9-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i64 1
// CHECK9-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK9-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK9:       arrayctor.cont:
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[_TMP2]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 1
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6:![0-9]+]]
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK9:       omp.inner.for.cond.cleanup:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
// CHECK9-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 8, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[IDXPROM4:%.*]] = sext i32 [[TMP11]] to i64
// CHECK9-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 [[IDXPROM4]]
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX5]], ptr align 4 [[TMP10]], i64 4, i1 false), !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK9-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK9-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN7]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done8:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102
// CHECK9-SAME: (i64 noundef [[I:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[I_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[I]], ptr [[I_ADDR]], align 8
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.omp_outlined, ptr [[I_ADDR]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.omp_outlined
// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[I_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store ptr [[I]], ptr [[I_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[I_ADDR]], align 8
// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12:![0-9]+]]
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[I1]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK9-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 2, ptr [[TMP0]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK9-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK9-NEXT:    [[VAR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK9-NEXT:    store i32 0, ptr [[T_VAR]], align 4
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 0
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 noundef signext 1)
// CHECK9-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYINIT_BEGIN]], i64 1
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef signext 2)
// CHECK9-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 8
// CHECK9-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 2, ptr [[TMP0]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 2, ptr [[TMP8]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP11]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK9-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 1, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.region_id, ptr [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK9:       omp_offload.failed:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49() #[[ATTR4]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done2:
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    ret i32 [[TMP16]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    store float 0.000000e+00, ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    store float [[TMP0]], ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef signext [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49
// CHECK9-SAME: () #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined)
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined
// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK9-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK9-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK9:       arrayctor.loop:
// CHECK9-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK9-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i64 1
// CHECK9-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK9-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK9:       arrayctor.cont:
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[_TMP2]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 1
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15:![0-9]+]]
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK9:       omp.inner.for.cond.cleanup:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
// CHECK9-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 8, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[IDXPROM4:%.*]] = sext i32 [[TMP11]] to i64
// CHECK9-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 [[IDXPROM4]]
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX5]], ptr align 4 [[TMP10]], i64 4, i1 false), !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK9-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK9-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN7]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done8:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    store i32 0, ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK9-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK11-NEXT:    [[G1:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK11-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK11-NEXT:    [[VAR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[KERNEL_ARGS3:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    store ptr [[G]], ptr [[G1]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK11-NEXT:    store i32 0, ptr [[T_VAR]], align 4
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const.main.vec, i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float noundef 1.000000e+00)
// CHECK11-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYINIT_BEGIN]], i32 1
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float noundef 2.000000e+00)
// CHECK11-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 4
// CHECK11-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 2, ptr [[TMP0]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK11-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr null, ptr [[TMP2]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr null, ptr [[TMP3]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK11-NEXT:    store ptr null, ptr [[TMP4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK11-NEXT:    store ptr null, ptr [[TMP5]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK11-NEXT:    store ptr null, ptr [[TMP6]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK11-NEXT:    store ptr null, ptr [[TMP7]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK11-NEXT:    store i64 2, ptr [[TMP8]], align 8
// CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK11-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK11-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP11]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK11-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93.region_id, ptr [[KERNEL_ARGS]])
// CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK11-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK11:       omp_offload.failed:
// CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93() #[[ATTR4:[0-9]+]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK11:       omp_offload.cont:
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4
// CHECK11-NEXT:    store i32 [[TMP15]], ptr [[I_CASTED]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I_CASTED]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP16]], ptr [[TMP17]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP16]], ptr [[TMP18]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK11-NEXT:    store ptr null, ptr [[TMP19]], align 4
// CHECK11-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 0
// CHECK11-NEXT:    store i32 2, ptr [[TMP22]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 1
// CHECK11-NEXT:    store i32 1, ptr [[TMP23]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[TMP20]], ptr [[TMP24]], align 4
// CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[TMP21]], ptr [[TMP25]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 4
// CHECK11-NEXT:    store ptr @.offload_sizes, ptr [[TMP26]], align 4
// CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 5
// CHECK11-NEXT:    store ptr @.offload_maptypes, ptr [[TMP27]], align 4
// CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 6
// CHECK11-NEXT:    store ptr null, ptr [[TMP28]], align 4
// CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 7
// CHECK11-NEXT:    store ptr null, ptr [[TMP29]], align 4
// CHECK11-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 8
// CHECK11-NEXT:    store i64 2, ptr [[TMP30]], align 8
// CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 9
// CHECK11-NEXT:    store i64 0, ptr [[TMP31]], align 8
// CHECK11-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 10
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP32]], align 4
// CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 11
// CHECK11-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP33]], align 4
// CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS3]], i32 0, i32 12
// CHECK11-NEXT:    store i32 0, ptr [[TMP34]], align 4
// CHECK11-NEXT:    [[TMP35:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 1, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.region_id, ptr [[KERNEL_ARGS3]])
// CHECK11-NEXT:    [[TMP36:%.*]] = icmp ne i32 [[TMP35]], 0
// CHECK11-NEXT:    br i1 [[TMP36]], label [[OMP_OFFLOAD_FAILED4:%.*]], label [[OMP_OFFLOAD_CONT5:%.*]]
// CHECK11:       omp_offload.failed4:
// CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102(i32 [[TMP16]]) #[[ATTR4]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT5]]
// CHECK11:       omp_offload.cont5:
// CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiET_v()
// CHECK11-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP37]], [[OMP_OFFLOAD_CONT5]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE6:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done6:
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK11-NEXT:    [[TMP38:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP38]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93
// CHECK11-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93.omp_outlined)
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l93.omp_outlined
// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK11-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK11-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[SVAR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK11:       arrayctor.loop:
// CHECK11-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK11-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i32 1
// CHECK11-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK11-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK11:       arrayctor.cont:
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[_TMP2]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 1
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP7:![0-9]+]]
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK11-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK11:       omp.inner.for.cond.cleanup:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
// CHECK11-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 [[TMP11]]
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP10]], i32 4, i1 false), !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK11-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK11-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK11:       .omp.final.then:
// CHECK11-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK11:       .omp.final.done:
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN6:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN6]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN6]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE7:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done7:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102
// CHECK11-SAME: (i32 noundef [[I:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[I_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[I]], ptr [[I_ADDR]], align 4
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.omp_outlined, ptr [[I_ADDR]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.omp_outlined
// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[I_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store ptr [[I]], ptr [[I_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[I_ADDR]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13:![0-9]+]]
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[I1]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK11-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK11-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK11:       .omp.final.then:
// CHECK11-NEXT:    store i32 2, ptr [[TMP0]], align 4
// CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK11:       .omp.final.done:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK11-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK11-NEXT:    [[VAR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK11-NEXT:    store i32 0, ptr [[T_VAR]], align 4
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 noundef 1)
// CHECK11-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYINIT_BEGIN]], i32 1
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef 2)
// CHECK11-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 4
// CHECK11-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 2, ptr [[TMP0]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK11-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr null, ptr [[TMP2]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr null, ptr [[TMP3]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK11-NEXT:    store ptr null, ptr [[TMP4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK11-NEXT:    store ptr null, ptr [[TMP5]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK11-NEXT:    store ptr null, ptr [[TMP6]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK11-NEXT:    store ptr null, ptr [[TMP7]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK11-NEXT:    store i64 2, ptr [[TMP8]], align 8
// CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK11-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK11-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP11]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK11-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 1, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.region_id, ptr [[KERNEL_ARGS]])
// CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK11-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK11:       omp_offload.failed:
// CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49() #[[ATTR4]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK11:       omp_offload.cont:
// CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done2:
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP16]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    store float 0.000000e+00, ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    store float [[TMP0]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49
// CHECK11-SAME: () #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined)
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined
// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK11-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK11-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK11:       arrayctor.loop:
// CHECK11-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK11-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i32 1
// CHECK11-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK11-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK11:       arrayctor.cont:
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[_TMP2]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 1
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16:![0-9]+]]
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK11-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK11:       omp.inner.for.cond.cleanup:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
// CHECK11-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 [[TMP11]]
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP10]], i32 4, i1 false), !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK11-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK11-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK11:       .omp.final.then:
// CHECK11-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK11:       .omp.final.done:
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN6]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN6]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE7:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done7:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    store i32 0, ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK11-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@main
// CHECK13-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK13-NEXT:    [[G1:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK13-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK13-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK13-NEXT:    [[VAR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[T_VAR2:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[VEC3:%.*]] = alloca [2 x i32], align 4
// CHECK13-NEXT:    [[S_ARR4:%.*]] = alloca [2 x %struct.S], align 4
// CHECK13-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK13-NEXT:    [[_TMP6:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[SVAR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I12:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB14:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB15:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IV16:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I17:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK13-NEXT:    store ptr [[G]], ptr [[G1]], align 8
// CHECK13-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK13-NEXT:    store i32 0, ptr [[T_VAR]], align 4
// CHECK13-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const.main.vec, i64 8, i1 false)
// CHECK13-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 0
// CHECK13-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float noundef 1.000000e+00)
// CHECK13-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYINIT_BEGIN]], i64 1
// CHECK13-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float noundef 2.000000e+00)
// CHECK13-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 8
// CHECK13-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i32 0, i32 0
// CHECK13-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK13-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK13:       arrayctor.loop:
// CHECK13-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK13-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK13-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i64 1
// CHECK13-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK13-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK13:       arrayctor.cont:
// CHECK13-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]])
// CHECK13-NEXT:    store ptr [[VAR5]], ptr [[_TMP6]], align 8
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2:![0-9]+]]
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK13:       omp.inner.for.cond.cleanup:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR2]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP5]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC3]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[_TMP6]], align 8, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP7]] to i64
// CHECK13-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i64 0, i64 [[IDXPROM7]]
// CHECK13-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX8]], ptr align 4 [[TMP6]], i64 4, i1 false), !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK13-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK13-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4:[0-9]+]]
// CHECK13-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN10]], i64 2
// CHECK13-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK13:       arraydestroy.body:
// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP9]], [[OMP_INNER_FOR_END]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK13-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK13-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK13:       arraydestroy.done11:
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB14]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_UB15]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_LB14]], align 4
// CHECK13-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV16]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND18:%.*]]
// CHECK13:       omp.inner.for.cond18:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV16]], align 4, !llvm.access.group [[ACC_GRP6:![0-9]+]]
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB15]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK13-NEXT:    [[CMP19:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    br i1 [[CMP19]], label [[OMP_INNER_FOR_BODY20:%.*]], label [[OMP_INNER_FOR_END26:%.*]]
// CHECK13:       omp.inner.for.body20:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV16]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK13-NEXT:    [[MUL21:%.*]] = mul nsw i32 [[TMP13]], 1
// CHECK13-NEXT:    [[ADD22:%.*]] = add nsw i32 0, [[MUL21]]
// CHECK13-NEXT:    store i32 [[ADD22]], ptr [[I17]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE23:%.*]]
// CHECK13:       omp.body.continue23:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC24:%.*]]
// CHECK13:       omp.inner.for.inc24:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV16]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK13-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK13-NEXT:    store i32 [[ADD25]], ptr [[DOTOMP_IV16]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND18]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK13:       omp.inner.for.end26:
// CHECK13-NEXT:    store i32 2, ptr [[I12]], align 4
// CHECK13-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v()
// CHECK13-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
// CHECK13-NEXT:    [[ARRAY_BEGIN27:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN27]], i64 2
// CHECK13-NEXT:    br label [[ARRAYDESTROY_BODY28:%.*]]
// CHECK13:       arraydestroy.body28:
// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST29:%.*]] = phi ptr [ [[TMP15]], [[OMP_INNER_FOR_END26]] ], [ [[ARRAYDESTROY_ELEMENT30:%.*]], [[ARRAYDESTROY_BODY28]] ]
// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENT30]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST29]], i64 -1
// CHECK13-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT30]]) #[[ATTR4]]
// CHECK13-NEXT:    [[ARRAYDESTROY_DONE31:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT30]], [[ARRAY_BEGIN27]]
// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE31]], label [[ARRAYDESTROY_DONE32:%.*]], label [[ARRAYDESTROY_BODY28]]
// CHECK13:       arraydestroy.done32:
// CHECK13-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK13-NEXT:    ret i32 [[TMP16]]
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK13-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK13-SAME: () #[[ATTR3:[0-9]+]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK13-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK13-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK13-NEXT:    [[VAR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[T_VAR2:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[VEC3:%.*]] = alloca [2 x i32], align 4
// CHECK13-NEXT:    [[S_ARR4:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK13-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK13-NEXT:    [[_TMP6:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK13-NEXT:    store i32 0, ptr [[T_VAR]], align 4
// CHECK13-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i64 8, i1 false)
// CHECK13-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 0
// CHECK13-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 noundef signext 1)
// CHECK13-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYINIT_BEGIN]], i64 1
// CHECK13-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef signext 2)
// CHECK13-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 8
// CHECK13-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 0
// CHECK13-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK13-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK13:       arrayctor.loop:
// CHECK13-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK13-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK13-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i64 1
// CHECK13-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK13-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK13:       arrayctor.cont:
// CHECK13-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]])
// CHECK13-NEXT:    store ptr [[VAR5]], ptr [[_TMP6]], align 8
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK13:       omp.inner.for.cond.cleanup:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR2]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP5]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC3]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[_TMP6]], align 8, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP7]] to i64
// CHECK13-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i64 0, i64 [[IDXPROM7]]
// CHECK13-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX8]], ptr align 4 [[TMP6]], i64 4, i1 false), !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK13-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK13-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4]]
// CHECK13-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN10]], i64 2
// CHECK13-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK13:       arraydestroy.body:
// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP9]], [[OMP_INNER_FOR_END]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK13-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK13-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK13:       arraydestroy.done11:
// CHECK13-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK13-NEXT:    [[ARRAY_BEGIN12:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN12]], i64 2
// CHECK13-NEXT:    br label [[ARRAYDESTROY_BODY13:%.*]]
// CHECK13:       arraydestroy.body13:
// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST14:%.*]] = phi ptr [ [[TMP10]], [[ARRAYDESTROY_DONE11]] ], [ [[ARRAYDESTROY_ELEMENT15:%.*]], [[ARRAYDESTROY_BODY13]] ]
// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENT15]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST14]], i64 -1
// CHECK13-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT15]]) #[[ATTR4]]
// CHECK13-NEXT:    [[ARRAYDESTROY_DONE16:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT15]], [[ARRAY_BEGIN12]]
// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE16]], label [[ARRAYDESTROY_DONE17:%.*]], label [[ARRAYDESTROY_BODY13]]
// CHECK13:       arraydestroy.done17:
// CHECK13-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK13-NEXT:    ret i32 [[TMP11]]
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK13-NEXT:    store float 0.000000e+00, ptr [[F]], align 4
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK13-NEXT:    store float [[TMP0]], ptr [[F]], align 4
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK13-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef signext [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK13-NEXT:    store i32 0, ptr [[F]], align 4
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@main
// CHECK15-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK15-NEXT:    [[G1:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK15-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK15-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK15-NEXT:    [[VAR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[T_VAR2:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VEC3:%.*]] = alloca [2 x i32], align 4
// CHECK15-NEXT:    [[S_ARR4:%.*]] = alloca [2 x %struct.S], align 4
// CHECK15-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK15-NEXT:    [[_TMP6:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[SVAR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I11:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[_TMP12:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB13:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB14:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IV15:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I16:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK15-NEXT:    store ptr [[G]], ptr [[G1]], align 4
// CHECK15-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK15-NEXT:    store i32 0, ptr [[T_VAR]], align 4
// CHECK15-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const.main.vec, i32 8, i1 false)
// CHECK15-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK15-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float noundef 1.000000e+00)
// CHECK15-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYINIT_BEGIN]], i32 1
// CHECK15-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float noundef 2.000000e+00)
// CHECK15-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 4
// CHECK15-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
// CHECK15-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i32 0, i32 0
// CHECK15-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK15-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK15:       arrayctor.loop:
// CHECK15-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK15-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK15-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i32 1
// CHECK15-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK15-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK15:       arrayctor.cont:
// CHECK15-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]])
// CHECK15-NEXT:    store ptr [[VAR5]], ptr [[_TMP6]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3:![0-9]+]]
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK15:       omp.inner.for.cond.cleanup:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR2]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC3]], i32 0, i32 [[TMP5]]
// CHECK15-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[_TMP6]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i32 0, i32 [[TMP7]]
// CHECK15-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX7]], ptr align 4 [[TMP6]], i32 4, i1 false), !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK15-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4:[0-9]+]]
// CHECK15-NEXT:    [[ARRAY_BEGIN9:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN9]], i32 2
// CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK15:       arraydestroy.body:
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP9]], [[OMP_INNER_FOR_END]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK15-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN9]]
// CHECK15-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE10:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK15:       arraydestroy.done10:
// CHECK15-NEXT:    store i32 0, ptr [[DOTOMP_LB13]], align 4
// CHECK15-NEXT:    store i32 1, ptr [[DOTOMP_UB14]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_LB13]], align 4
// CHECK15-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV15]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND17:%.*]]
// CHECK15:       omp.inner.for.cond17:
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV15]], align 4, !llvm.access.group [[ACC_GRP7:![0-9]+]]
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB14]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK15-NEXT:    [[CMP18:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK15-NEXT:    br i1 [[CMP18]], label [[OMP_INNER_FOR_BODY19:%.*]], label [[OMP_INNER_FOR_END25:%.*]]
// CHECK15:       omp.inner.for.body19:
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV15]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK15-NEXT:    [[MUL20:%.*]] = mul nsw i32 [[TMP13]], 1
// CHECK15-NEXT:    [[ADD21:%.*]] = add nsw i32 0, [[MUL20]]
// CHECK15-NEXT:    store i32 [[ADD21]], ptr [[I16]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE22:%.*]]
// CHECK15:       omp.body.continue22:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC23:%.*]]
// CHECK15:       omp.inner.for.inc23:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV15]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK15-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK15-NEXT:    store i32 [[ADD24]], ptr [[DOTOMP_IV15]], align 4, !llvm.access.group [[ACC_GRP7]]
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND17]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK15:       omp.inner.for.end25:
// CHECK15-NEXT:    store i32 2, ptr [[I11]], align 4
// CHECK15-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiET_v()
// CHECK15-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
// CHECK15-NEXT:    [[ARRAY_BEGIN26:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN26]], i32 2
// CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY27:%.*]]
// CHECK15:       arraydestroy.body27:
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST28:%.*]] = phi ptr [ [[TMP15]], [[OMP_INNER_FOR_END25]] ], [ [[ARRAYDESTROY_ELEMENT29:%.*]], [[ARRAYDESTROY_BODY27]] ]
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT29]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST28]], i32 -1
// CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT29]]) #[[ATTR4]]
// CHECK15-NEXT:    [[ARRAYDESTROY_DONE30:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT29]], [[ARRAY_BEGIN26]]
// CHECK15-NEXT:    br i1 [[ARRAYDESTROY_DONE30]], label [[ARRAYDESTROY_DONE31:%.*]], label [[ARRAYDESTROY_BODY27]]
// CHECK15:       arraydestroy.done31:
// CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK15-NEXT:    ret i32 [[TMP16]]
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK15-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK15-SAME: () #[[ATTR3:[0-9]+]] comdat {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK15-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK15-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK15-NEXT:    [[VAR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[T_VAR2:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VEC3:%.*]] = alloca [2 x i32], align 4
// CHECK15-NEXT:    [[S_ARR4:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK15-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK15-NEXT:    [[_TMP6:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK15-NEXT:    store i32 0, ptr [[T_VAR]], align 4
// CHECK15-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i32 8, i1 false)
// CHECK15-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK15-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 noundef 1)
// CHECK15-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYINIT_BEGIN]], i32 1
// CHECK15-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef 2)
// CHECK15-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 4
// CHECK15-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
// CHECK15-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 0
// CHECK15-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK15-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK15:       arrayctor.loop:
// CHECK15-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK15-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK15-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i32 1
// CHECK15-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK15-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK15:       arrayctor.cont:
// CHECK15-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]])
// CHECK15-NEXT:    store ptr [[VAR5]], ptr [[_TMP6]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10:![0-9]+]]
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK15:       omp.inner.for.cond.cleanup:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR2]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC3]], i32 0, i32 [[TMP5]]
// CHECK15-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[_TMP6]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 [[TMP7]]
// CHECK15-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX7]], ptr align 4 [[TMP6]], i32 4, i1 false), !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK15-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    store i32 2, ptr [[I]], align 4
// CHECK15-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4]]
// CHECK15-NEXT:    [[ARRAY_BEGIN9:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN9]], i32 2
// CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK15:       arraydestroy.body:
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP9]], [[OMP_INNER_FOR_END]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK15-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK15-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN9]]
// CHECK15-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE10:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK15:       arraydestroy.done10:
// CHECK15-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK15-NEXT:    [[ARRAY_BEGIN11:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN11]], i32 2
// CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY12:%.*]]
// CHECK15:       arraydestroy.body12:
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST13:%.*]] = phi ptr [ [[TMP10]], [[ARRAYDESTROY_DONE10]] ], [ [[ARRAYDESTROY_ELEMENT14:%.*]], [[ARRAYDESTROY_BODY12]] ]
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT14]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST13]], i32 -1
// CHECK15-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT14]]) #[[ATTR4]]
// CHECK15-NEXT:    [[ARRAYDESTROY_DONE15:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT14]], [[ARRAY_BEGIN11]]
// CHECK15-NEXT:    br i1 [[ARRAYDESTROY_DONE15]], label [[ARRAYDESTROY_DONE16:%.*]], label [[ARRAYDESTROY_BODY12]]
// CHECK15:       arraydestroy.done16:
// CHECK15-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK15-NEXT:    ret i32 [[TMP11]]
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK15-NEXT:    store float 0.000000e+00, ptr [[F]], align 4
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK15-NEXT:    store float [[TMP0]], ptr [[F]], align 4
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK15-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef [[TMP0]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK15-NEXT:    store i32 0, ptr [[F]], align 4
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK15-NEXT:    ret void
//