[clang] Implement lifetime analysis for lifetime_capture_by(X) (#115921)

[llvm-project.git] / clang / test / OpenMP / teams_firstprivate_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 _
// Test host codegen.
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -verify -Wno-vla -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -verify -Wno-vla -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK3

// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -verify -Wno-vla -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -verify -Wno-vla -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis -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 -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"

// RUN: %clang_cc1 -no-enable-noundef-analysis  -verify -Wno-vla -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK9
// RUN: %clang_cc1 -no-enable-noundef-analysis  -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis  -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK9
// RUN: %clang_cc1 -no-enable-noundef-analysis  -verify -Wno-vla -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK11
// RUN: %clang_cc1 -no-enable-noundef-analysis  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK11

// RUN: %clang_cc1 -no-enable-noundef-analysis  -verify -Wno-vla -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis  -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis  -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis  -verify -Wno-vla -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"

// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -verify -Wno-vla -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK17
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK17
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -verify -Wno-vla -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK19
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK19

// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -verify -Wno-vla -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -verify -Wno-vla -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
// RUN: %clang_cc1 -no-enable-noundef-analysis -DARRAY  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -include-pch %t -verify -Wno-vla %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// expected-no-diagnostics
#ifndef HEADER
#define HEADER
#ifndef ARRAY
struct St {
  int a, b;
  St() : a(0), b(0) {}
  St(const St &st) : a(st.a + st.b), b(0) {}
  ~St() {}
};

volatile int g __attribute__((aligned(128))) = 1212;

template <class T>
struct S {
  T f;
  S(T a) : f(a + g) {}
  S() : f(g) {}
  S(const S &s, St t = St()) : f(s.f + t.a) {}
  operator T() { return T(); }
  ~S() {}
};


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

int main() {
  static int sivar;
#ifdef LAMBDA
  [&]() {
  #pragma omp target
  #pragma omp teams firstprivate(g, sivar)
  {
    g = 1;
    sivar = 2;
    [&]() {
      g = 2;
      sivar = 4;
    }();
  }
  }();
  return 0;
#else
  S<float> test;
  int t_var = 0;
  int vec[] = {1, 2};
  S<float> s_arr[] = {1, 2};
  S<float> var(3);
  #pragma omp target
  #pragma omp teams firstprivate(t_var, vec, s_arr, var, sivar)
  {
    vec[0] = t_var;
    s_arr[0] = var;
    sivar = 2;
  }
  #pragma omp target
  #pragma omp teams firstprivate(t_var)
  {}
  return tmain<int>();
#endif
}













#else
struct St {
  int a, b;
  St() : a(0), b(0) {}
  St(const St &) { }
  ~St() {}
  void St_func(St s[2], int n, long double vla1[n]) {
    double vla2[n][n] __attribute__((aligned(128)));
    a = b;
    #pragma omp target
    #pragma omp teams firstprivate(s, vla1, vla2)
    vla1[b] = vla2[1][n - 1] = a = b;
  }
};

void array_func(float a[3], St s[2], int n, long double vla1[n]) {
  double vla2[n][n] __attribute__((aligned(128)));
  #pragma omp target
  #pragma omp teams firstprivate(a, s, vla1, vla2)
  s[0].St_func(s, n, vla1);
  ;
}

#endif
#endif
// CHECK1-LABEL: define {{[^@]+}}@main
// CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
// CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK1-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l91
// CHECK1-SAME: (i64 [[G:%.*]], i64 [[SIVAR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[G_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[G]], ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[G_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP0]], ptr [[G_CASTED]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[G_CASTED]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[SIVAR_CASTED]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1:[0-9]+]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l91.omp_outlined, i64 [[TMP1]], i64 [[TMP3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l91.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[G:%.*]], i64 [[SIVAR:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
// 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 i64 [[G]], ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
// CHECK1-NEXT:    store i32 1, ptr [[G_ADDR]], align 4
// CHECK1-NEXT:    store i32 2, ptr [[SIVAR_ADDR]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
// CHECK1-NEXT:    store ptr [[G_ADDR]], ptr [[TMP0]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
// CHECK1-NEXT:    store ptr [[SIVAR_ADDR]], ptr [[TMP1]], align 8
// CHECK1-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr nonnull align 8 dereferenceable(16) [[REF_TMP]])
// CHECK1-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
// CHECK3-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK3-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK3-NEXT:    ret i32 0
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l91
// CHECK3-SAME: (i32 [[G:%.*]], i32 [[SIVAR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[G_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[G]], ptr [[G_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[G_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[G_CASTED]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[G_CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[SIVAR_CASTED]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[SIVAR_CASTED]], align 4
// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1:[0-9]+]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l91.omp_outlined, i32 [[TMP1]], i32 [[TMP3]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l91.omp_outlined
// CHECK3-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i32 [[G:%.*]], i32 [[SIVAR:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = 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 i32 [[G]], ptr [[G_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
// CHECK3-NEXT:    store i32 1, ptr [[G_ADDR]], align 4
// CHECK3-NEXT:    store i32 2, ptr [[SIVAR_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
// CHECK3-NEXT:    store ptr [[G_ADDR]], ptr [[TMP0]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
// CHECK3-NEXT:    store ptr [[SIVAR_ADDR]], ptr [[TMP1]], align 4
// CHECK3-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr nonnull align 4 dereferenceable(8) [[REF_TMP]])
// CHECK3-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// 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 [[STRUCT_S]], align 4
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 8
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    [[T_VAR_CASTED1:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS2:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS3:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS4:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ev(ptr 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:    call void @_ZN1SIfEC1Ef(ptr nonnull align 4 dereferenceable(4) [[S_ARR]], float 1.000000e+00)
// CHECK9-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[S_ARR]], i64 1
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(ptr nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float 2.000000e+00)
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(ptr nonnull align 4 dereferenceable(4) [[VAR]], float 3.000000e+00)
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[T_VAR]], align 4
// CHECK9-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr @_ZZ4mainE5sivar, align 4
// CHECK9-NEXT:    store i32 [[TMP2]], ptr [[SIVAR_CASTED]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP4]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP5]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK9-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK9-NEXT:    store ptr [[VEC]], ptr [[TMP7]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK9-NEXT:    store ptr [[VEC]], ptr [[TMP8]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK9-NEXT:    store ptr null, ptr [[TMP9]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[S_ARR]], ptr [[TMP10]], align 8
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[S_ARR]], ptr [[TMP11]], align 8
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP12]], align 8
// CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[TMP13]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[TMP14]], align 8
// CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP15]], align 8
// CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK9-NEXT:    store i64 [[TMP3]], ptr [[TMP16]], align 8
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK9-NEXT:    store i64 [[TMP3]], ptr [[TMP17]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK9-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP21]], align 4
// CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 5, ptr [[TMP22]], align 4
// CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[TMP19]], ptr [[TMP23]], align 8
// CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[TMP20]], ptr [[TMP24]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store ptr @.offload_sizes, ptr [[TMP25]], align 8
// CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP26]], align 8
// CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP27]], align 8
// CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP28]], align 8
// CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 0, ptr [[TMP29]], align 8
// CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP30]], align 8
// CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP31]], align 4
// CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP32]], align 4
// CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP33]], align 4
// CHECK9-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l109.region_id, ptr [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK9-NEXT:    br i1 [[TMP35]], 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_l109(i64 [[TMP1]], ptr [[VEC]], ptr [[S_ARR]], ptr [[VAR]], i64 [[TMP3]]) #[[ATTR4:[0-9]+]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    [[TMP36:%.*]] = load i32, ptr [[T_VAR]], align 4
// CHECK9-NEXT:    store i32 [[TMP36]], ptr [[T_VAR_CASTED1]], align 4
// CHECK9-NEXT:    [[TMP37:%.*]] = load i64, ptr [[T_VAR_CASTED1]], align 8
// CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP37]], ptr [[TMP38]], align 8
// CHECK9-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP37]], ptr [[TMP39]], align 8
// CHECK9-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS4]], i64 0, i64 0
// CHECK9-NEXT:    store ptr null, ptr [[TMP40]], align 8
// CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP43]], align 4
// CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK9-NEXT:    store i32 1, ptr [[TMP44]], align 4
// CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[TMP41]], ptr [[TMP45]], align 8
// CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[TMP42]], ptr [[TMP46]], align 8
// CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK9-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP47]], align 8
// CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK9-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP48]], align 8
// CHECK9-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP49]], align 8
// CHECK9-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP50]], align 8
// CHECK9-NEXT:    [[TMP51:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK9-NEXT:    store i64 0, ptr [[TMP51]], align 8
// CHECK9-NEXT:    [[TMP52:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP52]], align 8
// CHECK9-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP53]], align 4
// CHECK9-NEXT:    [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP54]], align 4
// CHECK9-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP55]], align 4
// CHECK9-NEXT:    [[TMP56:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116.region_id, ptr [[KERNEL_ARGS5]])
// CHECK9-NEXT:    [[TMP57:%.*]] = icmp ne i32 [[TMP56]], 0
// CHECK9-NEXT:    br i1 [[TMP57]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK9:       omp_offload.failed6:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116(i64 [[TMP37]]) #[[ATTR4]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK9:       omp_offload.cont7:
// CHECK9-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
// CHECK9-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP58]], [[OMP_OFFLOAD_CONT7]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr 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_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done8:
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK9-NEXT:    [[TMP59:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    ret i32 [[TMP59]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK9-SAME: (ptr 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 nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[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 nonnull align 4 dereferenceable(4) [[THIS1]], float [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l109
// CHECK9-SAME: (i64 [[T_VAR:%.*]], ptr nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 [[SIVAR:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK9-NEXT:    store i32 [[TMP3]], ptr [[T_VAR_CASTED]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
// CHECK9-NEXT:    store i32 [[TMP5]], ptr [[SIVAR_CASTED]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l109.omp_outlined, ptr [[TMP0]], i64 [[TMP4]], ptr [[TMP1]], ptr [[TMP2]], i64 [[TMP6]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l109.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 [[T_VAR:%.*]], ptr nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 [[SIVAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VEC1:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR2:%.*]] = alloca [2 x %struct.S], align 4
// CHECK9-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
// CHECK9-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK9-NEXT:    [[AGG_TMP5:%.*]] = alloca [[STRUCT_ST]], 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 [[VEC]], ptr [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC1]], ptr align 4 [[TMP0]], i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK9:       omp.arraycpy.body:
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    call void @_ZN2StC1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP]])
// CHECK9-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr [[AGG_TMP]])
// CHECK9-NEXT:    call void @_ZN2StD1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR4]]
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
// CHECK9:       omp.arraycpy.done3:
// CHECK9-NEXT:    call void @_ZN2StC1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP5]])
// CHECK9-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[VAR4]], ptr nonnull align 4 dereferenceable(4) [[TMP2]], ptr [[AGG_TMP5]])
// CHECK9-NEXT:    call void @_ZN2StD1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP5]]) #[[ATTR4]]
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC1]], i64 0, i64 0
// CHECK9-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i64 0, i64 0
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX6]], ptr align 4 [[VAR4]], i64 4, i1 false)
// CHECK9-NEXT:    store i32 2, ptr [[SIVAR_ADDR]], align 4
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN7]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[OMP_ARRAYCPY_DONE3]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr 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 {{[^@]+}}@_ZN2StC1Ev
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(8) [[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 @_ZN2StC2Ev(ptr nonnull align 4 dereferenceable(8) [[THIS1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1ERKS0_2St
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr nonnull align 4 dereferenceable(4) [[S:%.*]], ptr [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIfEC2ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[THIS1]], ptr nonnull align 4 dereferenceable(4) [[TMP0]], ptr [[T]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN2StD1Ev
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(8) [[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 @_ZN2StD2Ev(ptr nonnull align 4 dereferenceable(8) [[THIS1]]) #[[ATTR4]]
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK9-SAME: (ptr 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 nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116
// CHECK9-SAME: (i64 [[T_VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK9-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116.omp_outlined, i64 [[TMP1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[T_VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// 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 i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK9-SAME: () #[[ATTR1]] 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 128
// CHECK9-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 128
// CHECK9-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 128
// CHECK9-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0]], align 128
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 8
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    [[T_VAR_CASTED1:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS2:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS3:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS4:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ev(ptr nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK9-NEXT:    store i32 0, ptr [[T_VAR]], align 128
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[VEC]], ptr align 128 @__const._Z5tmainIiET_v.vec, i64 8, i1 false)
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ei(ptr nonnull align 4 dereferenceable(4) [[S_ARR]], i32 signext 1)
// CHECK9-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[S_ARR]], i64 1
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ei(ptr nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 signext 2)
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ei(ptr nonnull align 4 dereferenceable(4) [[VAR]], i32 signext 3)
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[T_VAR]], align 128
// CHECK9-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP2]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP3]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK9-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK9-NEXT:    store ptr [[VEC]], ptr [[TMP5]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK9-NEXT:    store ptr [[VEC]], ptr [[TMP6]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK9-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[S_ARR]], ptr [[TMP8]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[S_ARR]], ptr [[TMP9]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP10]], align 8
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[TMP11]], align 8
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[TMP12]], align 8
// CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP13]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP16]], align 4
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 4, ptr [[TMP17]], align 4
// CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[TMP14]], ptr [[TMP18]], align 8
// CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[TMP15]], ptr [[TMP19]], align 8
// CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP20]], align 8
// CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP21]], align 8
// CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP23]], align 8
// CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP25]], align 8
// CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP27]], align 4
// CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP28]], align 4
// CHECK9-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l75.region_id, ptr [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK9-NEXT:    br i1 [[TMP30]], 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_l75(i64 [[TMP1]], ptr [[VEC]], ptr [[S_ARR]], ptr [[VAR]]) #[[ATTR4]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    [[TMP31:%.*]] = load i32, ptr [[T_VAR]], align 128
// CHECK9-NEXT:    store i32 [[TMP31]], ptr [[T_VAR_CASTED1]], align 4
// CHECK9-NEXT:    [[TMP32:%.*]] = load i64, ptr [[T_VAR_CASTED1]], align 8
// CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP32]], ptr [[TMP33]], align 8
// CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP32]], ptr [[TMP34]], align 8
// CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS4]], i64 0, i64 0
// CHECK9-NEXT:    store ptr null, ptr [[TMP35]], align 8
// CHECK9-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP38]], align 4
// CHECK9-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK9-NEXT:    store i32 1, ptr [[TMP39]], align 4
// CHECK9-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[TMP36]], ptr [[TMP40]], align 8
// CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[TMP37]], ptr [[TMP41]], align 8
// CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK9-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP42]], align 8
// CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK9-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP43]], align 8
// CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP44]], align 8
// CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP45]], align 8
// CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK9-NEXT:    store i64 0, ptr [[TMP46]], align 8
// CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP47]], align 8
// CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP48]], align 4
// CHECK9-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP49]], align 4
// CHECK9-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP50]], align 4
// CHECK9-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81.region_id, ptr [[KERNEL_ARGS5]])
// CHECK9-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
// CHECK9-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK9:       omp_offload.failed6:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81(i64 [[TMP32]]) #[[ATTR4]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK9:       omp_offload.cont7:
// CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP53:%.*]] = 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 [ [[TMP53]], [[OMP_OFFLOAD_CONT7]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr 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_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done8:
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK9-NEXT:    [[TMP54:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    ret i32 [[TMP54]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK9-SAME: (ptr 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 nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 128
// CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
// CHECK9-NEXT:    store float [[CONV]], ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[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 nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 128
// CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
// CHECK9-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
// CHECK9-NEXT:    store float [[ADD]], ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN2StC2Ev
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(8) [[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:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    store i32 0, ptr [[A]], align 4
// CHECK9-NEXT:    [[B:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 1
// CHECK9-NEXT:    store i32 0, ptr [[B]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2ERKS0_2St
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr nonnull align 4 dereferenceable(4) [[S:%.*]], ptr [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK9-NEXT:    [[F2:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[TMP0]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP1:%.*]] = load float, ptr [[F2]], align 4
// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[T]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A]], align 4
// CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP2]] to float
// CHECK9-NEXT:    [[ADD:%.*]] = fadd float [[TMP1]], [[CONV]]
// CHECK9-NEXT:    store float [[ADD]], ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN2StD2Ev
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(8) [[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 {{[^@]+}}@_ZN1SIfED2Ev
// CHECK9-SAME: (ptr 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 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 nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 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 nonnull align 4 dereferenceable(4) [[THIS1]], i32 signext [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l75
// CHECK9-SAME: (i64 [[T_VAR:%.*]], ptr nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK9-NEXT:    store i32 [[TMP3]], ptr [[T_VAR_CASTED]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l75.omp_outlined, ptr [[TMP0]], i64 [[TMP4]], ptr [[TMP1]], ptr [[TMP2]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l75.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 [[T_VAR:%.*]], ptr nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[VEC1:%.*]] = alloca [2 x i32], align 128
// CHECK9-NEXT:    [[S_ARR2:%.*]] = alloca [2 x %struct.S.0], align 128
// CHECK9-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
// CHECK9-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S_0:%.*]], align 128
// CHECK9-NEXT:    [[AGG_TMP5:%.*]] = alloca [[STRUCT_ST]], 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 [[VEC]], ptr [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[VEC1]], ptr align 128 [[TMP0]], i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR2]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK9:       omp.arraycpy.body:
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    call void @_ZN2StC1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP]])
// CHECK9-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr [[AGG_TMP]])
// CHECK9-NEXT:    call void @_ZN2StD1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR4]]
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
// CHECK9:       omp.arraycpy.done3:
// CHECK9-NEXT:    call void @_ZN2StC1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP5]])
// CHECK9-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[VAR4]], ptr nonnull align 4 dereferenceable(4) [[TMP2]], ptr [[AGG_TMP5]])
// CHECK9-NEXT:    call void @_ZN2StD1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP5]]) #[[ATTR4]]
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC1]], i64 0, i64 0
// CHECK9-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 128
// CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR2]], i64 0, i64 0
// CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[ARRAYIDX6]], ptr align 128 [[VAR4]], i64 4, i1 false)
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR2]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP5:%.*]] = 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 [ [[TMP5]], [[OMP_ARRAYCPY_DONE3]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(ptr 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 {{[^@]+}}@_ZN1SIiEC1ERKS0_2St
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr nonnull align 4 dereferenceable(4) [[S:%.*]], ptr [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIiEC2ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[THIS1]], ptr nonnull align 4 dereferenceable(4) [[TMP0]], ptr [[T]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK9-SAME: (ptr 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 nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81
// CHECK9-SAME: (i64 [[T_VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK9-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81.omp_outlined, i64 [[TMP1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[T_VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// 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 i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK9-SAME: (ptr 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 nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 128
// CHECK9-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 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 nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 128
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], [[TMP1]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC2ERKS0_2St
// CHECK9-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr nonnull align 4 dereferenceable(4) [[S:%.*]], ptr [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK9-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK9-NEXT:    [[F2:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0]], ptr [[TMP0]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[F2]], align 4
// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[T]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A]], align 4
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[TMP2]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK9-SAME: (ptr 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
//
//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK11-NEXT:    [[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:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 4
// CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK11-NEXT:    [[T_VAR_CASTED1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS2:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS3:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS4:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ev(ptr 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:    call void @_ZN1SIfEC1Ef(ptr nonnull align 4 dereferenceable(4) [[S_ARR]], float 1.000000e+00)
// CHECK11-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[S_ARR]], i32 1
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float 2.000000e+00)
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr nonnull align 4 dereferenceable(4) [[VAR]], float 3.000000e+00)
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[T_VAR]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr @_ZZ4mainE5sivar, align 4
// CHECK11-NEXT:    store i32 [[TMP2]], ptr [[SIVAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[SIVAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP5]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK11-NEXT:    store ptr null, ptr [[TMP6]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK11-NEXT:    store ptr [[VEC]], ptr [[TMP7]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK11-NEXT:    store ptr [[VEC]], ptr [[TMP8]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK11-NEXT:    store ptr null, ptr [[TMP9]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[TMP10]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[TMP11]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr null, ptr [[TMP12]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[TMP13]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[TMP14]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr null, ptr [[TMP15]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK11-NEXT:    store i32 [[TMP3]], ptr [[TMP16]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK11-NEXT:    store i32 [[TMP3]], ptr [[TMP17]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK11-NEXT:    store ptr null, ptr [[TMP18]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 3, ptr [[TMP21]], align 4
// CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK11-NEXT:    store i32 5, ptr [[TMP22]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[TMP19]], ptr [[TMP23]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[TMP20]], ptr [[TMP24]], align 4
// CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK11-NEXT:    store ptr @.offload_sizes, ptr [[TMP25]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK11-NEXT:    store ptr @.offload_maptypes, ptr [[TMP26]], align 4
// CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK11-NEXT:    store ptr null, ptr [[TMP27]], align 4
// CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK11-NEXT:    store ptr null, ptr [[TMP28]], align 4
// CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK11-NEXT:    store i64 0, ptr [[TMP29]], align 8
// CHECK11-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK11-NEXT:    store i64 0, ptr [[TMP30]], align 8
// CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP31]], align 4
// CHECK11-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP32]], align 4
// CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK11-NEXT:    store i32 0, ptr [[TMP33]], align 4
// CHECK11-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l109.region_id, ptr [[KERNEL_ARGS]])
// CHECK11-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK11-NEXT:    br i1 [[TMP35]], 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_l109(i32 [[TMP1]], ptr [[VEC]], ptr [[S_ARR]], ptr [[VAR]], i32 [[TMP3]]) #[[ATTR4:[0-9]+]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK11:       omp_offload.cont:
// CHECK11-NEXT:    [[TMP36:%.*]] = load i32, ptr [[T_VAR]], align 4
// CHECK11-NEXT:    store i32 [[TMP36]], ptr [[T_VAR_CASTED1]], align 4
// CHECK11-NEXT:    [[TMP37:%.*]] = load i32, ptr [[T_VAR_CASTED1]], align 4
// CHECK11-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP37]], ptr [[TMP38]], align 4
// CHECK11-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP37]], ptr [[TMP39]], align 4
// CHECK11-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS4]], i32 0, i32 0
// CHECK11-NEXT:    store ptr null, ptr [[TMP40]], align 4
// CHECK11-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK11-NEXT:    store i32 3, ptr [[TMP43]], align 4
// CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK11-NEXT:    store i32 1, ptr [[TMP44]], align 4
// CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[TMP41]], ptr [[TMP45]], align 4
// CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[TMP42]], ptr [[TMP46]], align 4
// CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK11-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP47]], align 4
// CHECK11-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK11-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP48]], align 4
// CHECK11-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK11-NEXT:    store ptr null, ptr [[TMP49]], align 4
// CHECK11-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK11-NEXT:    store ptr null, ptr [[TMP50]], align 4
// CHECK11-NEXT:    [[TMP51:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK11-NEXT:    store i64 0, ptr [[TMP51]], align 8
// CHECK11-NEXT:    [[TMP52:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
// CHECK11-NEXT:    store i64 0, ptr [[TMP52]], align 8
// CHECK11-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP53]], align 4
// CHECK11-NEXT:    [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP54]], align 4
// CHECK11-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
// CHECK11-NEXT:    store i32 0, ptr [[TMP55]], align 4
// CHECK11-NEXT:    [[TMP56:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116.region_id, ptr [[KERNEL_ARGS5]])
// CHECK11-NEXT:    [[TMP57:%.*]] = icmp ne i32 [[TMP56]], 0
// CHECK11-NEXT:    br i1 [[TMP57]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK11:       omp_offload.failed6:
// CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116(i32 [[TMP37]]) #[[ATTR4]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK11:       omp_offload.cont7:
// CHECK11-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiET_v()
// CHECK11-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP58]], [[OMP_OFFLOAD_CONT7]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr 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_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done8:
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK11-NEXT:    [[TMP59:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP59]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK11-SAME: (ptr 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 nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[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 nonnull align 4 dereferenceable(4) [[THIS1]], float [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l109
// CHECK11-SAME: (i32 [[T_VAR:%.*]], ptr nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 [[SIVAR:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], ptr [[SIVAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[SIVAR_CASTED]], align 4
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l109.omp_outlined, ptr [[TMP0]], i32 [[TMP4]], ptr [[TMP1]], ptr [[TMP2]], i32 [[TMP6]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l109.omp_outlined
// CHECK11-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 [[T_VAR:%.*]], ptr nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 [[SIVAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC1:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR2:%.*]] = alloca [2 x %struct.S], align 4
// CHECK11-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
// CHECK11-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK11-NEXT:    [[AGG_TMP5:%.*]] = alloca [[STRUCT_ST]], 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 [[VEC]], ptr [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC1]], ptr align 4 [[TMP0]], i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK11:       omp.arraycpy.body:
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    call void @_ZN2StC1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP]])
// CHECK11-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr [[AGG_TMP]])
// CHECK11-NEXT:    call void @_ZN2StD1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR4]]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
// CHECK11:       omp.arraycpy.done3:
// CHECK11-NEXT:    call void @_ZN2StC1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP5]])
// CHECK11-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[VAR4]], ptr nonnull align 4 dereferenceable(4) [[TMP2]], ptr [[AGG_TMP5]])
// CHECK11-NEXT:    call void @_ZN2StD1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP5]]) #[[ATTR4]]
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC1]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX6]], ptr align 4 [[VAR4]], i32 4, i1 false)
// CHECK11-NEXT:    store i32 2, ptr [[SIVAR_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN7]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[OMP_ARRAYCPY_DONE3]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done8:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN2StC1Ev
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(8) [[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 @_ZN2StC2Ev(ptr nonnull align 4 dereferenceable(8) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1ERKS0_2St
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr nonnull align 4 dereferenceable(4) [[S:%.*]], ptr [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC2ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[THIS1]], ptr nonnull align 4 dereferenceable(4) [[TMP0]], ptr [[T]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN2StD1Ev
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(8) [[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 @_ZN2StD2Ev(ptr nonnull align 4 dereferenceable(8) [[THIS1]]) #[[ATTR4]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK11-SAME: (ptr 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 nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116
// CHECK11-SAME: (i32 [[T_VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116.omp_outlined, i32 [[TMP1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116.omp_outlined
// CHECK11-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i32 [[T_VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = 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 i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK11-SAME: () #[[ATTR1]] 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 128
// CHECK11-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 128
// CHECK11-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 128
// CHECK11-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0]], align 128
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 4
// CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK11-NEXT:    [[T_VAR_CASTED1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS2:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS3:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS4:%.*]] = alloca [1 x ptr], align 4
// CHECK11-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ev(ptr nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK11-NEXT:    store i32 0, ptr [[T_VAR]], align 128
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 128 [[VEC]], ptr align 128 @__const._Z5tmainIiET_v.vec, i32 8, i1 false)
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(ptr nonnull align 4 dereferenceable(4) [[S_ARR]], i32 1)
// CHECK11-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[S_ARR]], i32 1
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(ptr nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 2)
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(ptr nonnull align 4 dereferenceable(4) [[VAR]], i32 3)
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[T_VAR]], align 128
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP2]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP3]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK11-NEXT:    store ptr null, ptr [[TMP4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK11-NEXT:    store ptr [[VEC]], ptr [[TMP5]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK11-NEXT:    store ptr [[VEC]], ptr [[TMP6]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK11-NEXT:    store ptr null, ptr [[TMP7]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[TMP8]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[TMP9]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr null, ptr [[TMP10]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[TMP11]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[TMP12]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr null, ptr [[TMP13]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 3, ptr [[TMP16]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK11-NEXT:    store i32 4, ptr [[TMP17]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[TMP14]], ptr [[TMP18]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[TMP15]], ptr [[TMP19]], align 4
// CHECK11-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK11-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP20]], align 4
// CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK11-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP21]], align 4
// CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK11-NEXT:    store ptr null, ptr [[TMP22]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK11-NEXT:    store ptr null, ptr [[TMP23]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK11-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK11-NEXT:    store i64 0, ptr [[TMP25]], align 8
// CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP27]], align 4
// CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK11-NEXT:    store i32 0, ptr [[TMP28]], align 4
// CHECK11-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l75.region_id, ptr [[KERNEL_ARGS]])
// CHECK11-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK11-NEXT:    br i1 [[TMP30]], 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_l75(i32 [[TMP1]], ptr [[VEC]], ptr [[S_ARR]], ptr [[VAR]]) #[[ATTR4]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK11:       omp_offload.cont:
// CHECK11-NEXT:    [[TMP31:%.*]] = load i32, ptr [[T_VAR]], align 128
// CHECK11-NEXT:    store i32 [[TMP31]], ptr [[T_VAR_CASTED1]], align 4
// CHECK11-NEXT:    [[TMP32:%.*]] = load i32, ptr [[T_VAR_CASTED1]], align 4
// CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP32]], ptr [[TMP33]], align 4
// CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP32]], ptr [[TMP34]], align 4
// CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS4]], i32 0, i32 0
// CHECK11-NEXT:    store ptr null, ptr [[TMP35]], align 4
// CHECK11-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK11-NEXT:    store i32 3, ptr [[TMP38]], align 4
// CHECK11-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK11-NEXT:    store i32 1, ptr [[TMP39]], align 4
// CHECK11-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK11-NEXT:    store ptr [[TMP36]], ptr [[TMP40]], align 4
// CHECK11-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK11-NEXT:    store ptr [[TMP37]], ptr [[TMP41]], align 4
// CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK11-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP42]], align 4
// CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK11-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP43]], align 4
// CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK11-NEXT:    store ptr null, ptr [[TMP44]], align 4
// CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK11-NEXT:    store ptr null, ptr [[TMP45]], align 4
// CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK11-NEXT:    store i64 0, ptr [[TMP46]], align 8
// CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
// CHECK11-NEXT:    store i64 0, ptr [[TMP47]], align 8
// CHECK11-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP48]], align 4
// CHECK11-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
// CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP49]], align 4
// CHECK11-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
// CHECK11-NEXT:    store i32 0, ptr [[TMP50]], align 4
// CHECK11-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81.region_id, ptr [[KERNEL_ARGS5]])
// CHECK11-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
// CHECK11-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK11:       omp_offload.failed6:
// CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81(i32 [[TMP32]]) #[[ATTR4]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK11:       omp_offload.cont7:
// CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP53:%.*]] = 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 [ [[TMP53]], [[OMP_OFFLOAD_CONT7]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr 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_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done8:
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK11-NEXT:    [[TMP54:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP54]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK11-SAME: (ptr 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 nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 128
// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
// CHECK11-NEXT:    store float [[CONV]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[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 nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 128
// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
// CHECK11-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
// CHECK11-NEXT:    store float [[ADD]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN2StC2Ev
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(8) [[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:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    store i32 0, ptr [[A]], align 4
// CHECK11-NEXT:    [[B:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 1
// CHECK11-NEXT:    store i32 0, ptr [[B]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2ERKS0_2St
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr nonnull align 4 dereferenceable(4) [[S:%.*]], ptr [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK11-NEXT:    [[F2:%.*]] = getelementptr inbounds nuw [[STRUCT_S]], ptr [[TMP0]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP1:%.*]] = load float, ptr [[F2]], align 4
// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[T]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP2]] to float
// CHECK11-NEXT:    [[ADD:%.*]] = fadd float [[TMP1]], [[CONV]]
// CHECK11-NEXT:    store float [[ADD]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN2StD2Ev
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(8) [[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 {{[^@]+}}@_ZN1SIfED2Ev
// CHECK11-SAME: (ptr 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 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 nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[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 nonnull align 4 dereferenceable(4) [[THIS1]], i32 [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l75
// CHECK11-SAME: (i32 [[T_VAR:%.*]], ptr nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l75.omp_outlined, ptr [[TMP0]], i32 [[TMP4]], ptr [[TMP1]], ptr [[TMP2]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l75.omp_outlined
// CHECK11-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 [[T_VAR:%.*]], ptr nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[VEC1:%.*]] = alloca [2 x i32], align 128
// CHECK11-NEXT:    [[S_ARR2:%.*]] = alloca [2 x %struct.S.0], align 128
// CHECK11-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
// CHECK11-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S_0:%.*]], align 128
// CHECK11-NEXT:    [[AGG_TMP5:%.*]] = alloca [[STRUCT_ST]], 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 [[VEC]], ptr [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 128 [[VEC1]], ptr align 128 [[TMP0]], i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR2]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK11:       omp.arraycpy.body:
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    call void @_ZN2StC1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP]])
// CHECK11-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr [[AGG_TMP]])
// CHECK11-NEXT:    call void @_ZN2StD1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR4]]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
// CHECK11:       omp.arraycpy.done3:
// CHECK11-NEXT:    call void @_ZN2StC1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP5]])
// CHECK11-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[VAR4]], ptr nonnull align 4 dereferenceable(4) [[TMP2]], ptr [[AGG_TMP5]])
// CHECK11-NEXT:    call void @_ZN2StD1Ev(ptr nonnull align 4 dereferenceable(8) [[AGG_TMP5]]) #[[ATTR4]]
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC1]], i32 0, i32 0
// CHECK11-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 128
// CHECK11-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR2]], i32 0, i32 0
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 128 [[ARRAYIDX6]], ptr align 128 [[VAR4]], i32 4, i1 false)
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR2]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN7]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[OMP_ARRAYCPY_DONE3]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done8:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC1ERKS0_2St
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr nonnull align 4 dereferenceable(4) [[S:%.*]], ptr [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiEC2ERKS0_2St(ptr nonnull align 4 dereferenceable(4) [[THIS1]], ptr nonnull align 4 dereferenceable(4) [[TMP0]], ptr [[T]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK11-SAME: (ptr 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 nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81
// CHECK11-SAME: (i32 [[T_VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81.omp_outlined, i32 [[TMP1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l81.omp_outlined
// CHECK11-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i32 [[T_VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = 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 i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK11-SAME: (ptr 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 nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 128
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[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 nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 128
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], [[TMP1]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2ERKS0_2St
// CHECK11-SAME: (ptr nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr nonnull align 4 dereferenceable(4) [[S:%.*]], ptr [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK11-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK11-NEXT:    [[F2:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0]], ptr [[TMP0]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[F2]], align 4
// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[T]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[TMP2]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK11-SAME: (ptr 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
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z10array_funcPfP2StiPg
// CHECK17-SAME: (ptr [[A:%.*]], ptr [[S:%.*]], i32 signext [[N:%.*]], ptr [[VLA1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [8 x ptr], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [8 x ptr], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [8 x ptr], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [8 x i64], align 8
// CHECK17-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK17-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
// CHECK17-NEXT:    [[TMP6:%.*]] = call ptr @llvm.stacksave.p0()
// CHECK17-NEXT:    store ptr [[TMP6]], ptr [[SAVED_STACK]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = mul nuw i64 [[TMP3]], [[TMP5]]
// CHECK17-NEXT:    [[VLA:%.*]] = alloca double, i64 [[TMP7]], align 128
// CHECK17-NEXT:    store i64 [[TMP3]], ptr [[__VLA_EXPR0]], align 8
// CHECK17-NEXT:    store i64 [[TMP5]], ptr [[__VLA_EXPR1]], align 8
// CHECK17-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP11]], ptr [[N_CASTED]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK17-NEXT:    [[TMP13:%.*]] = mul nuw i64 [[TMP3]], [[TMP5]]
// CHECK17-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 8
// CHECK17-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes, i64 64, i1 false)
// CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    store ptr [[TMP8]], ptr [[TMP15]], align 8
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    store ptr [[TMP8]], ptr [[TMP16]], align 8
// CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    store ptr [[TMP9]], ptr [[TMP18]], align 8
// CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    store ptr [[TMP9]], ptr [[TMP19]], align 8
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK17-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP1]], ptr [[TMP21]], align 8
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP1]], ptr [[TMP22]], align 8
// CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK17-NEXT:    store ptr null, ptr [[TMP23]], align 8
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK17-NEXT:    store ptr [[TMP10]], ptr [[TMP24]], align 8
// CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK17-NEXT:    store ptr [[TMP10]], ptr [[TMP25]], align 8
// CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK17-NEXT:    store ptr null, ptr [[TMP26]], align 8
// CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK17-NEXT:    store i64 [[TMP3]], ptr [[TMP27]], align 8
// CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK17-NEXT:    store i64 [[TMP3]], ptr [[TMP28]], align 8
// CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK17-NEXT:    store ptr null, ptr [[TMP29]], align 8
// CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK17-NEXT:    store i64 [[TMP5]], ptr [[TMP30]], align 8
// CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK17-NEXT:    store i64 [[TMP5]], ptr [[TMP31]], align 8
// CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
// CHECK17-NEXT:    store ptr null, ptr [[TMP32]], align 8
// CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK17-NEXT:    store ptr [[VLA]], ptr [[TMP33]], align 8
// CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK17-NEXT:    store ptr [[VLA]], ptr [[TMP34]], align 8
// CHECK17-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [8 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK17-NEXT:    store i64 [[TMP14]], ptr [[TMP35]], align 8
// CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 6
// CHECK17-NEXT:    store ptr null, ptr [[TMP36]], align 8
// CHECK17-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK17-NEXT:    store i64 [[TMP12]], ptr [[TMP37]], align 8
// CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK17-NEXT:    store i64 [[TMP12]], ptr [[TMP38]], align 8
// CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 7
// CHECK17-NEXT:    store ptr null, ptr [[TMP39]], align 8
// CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [8 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK17-NEXT:    store i32 3, ptr [[TMP43]], align 4
// CHECK17-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK17-NEXT:    store i32 8, ptr [[TMP44]], align 4
// CHECK17-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK17-NEXT:    store ptr [[TMP40]], ptr [[TMP45]], align 8
// CHECK17-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK17-NEXT:    store ptr [[TMP41]], ptr [[TMP46]], align 8
// CHECK17-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK17-NEXT:    store ptr [[TMP42]], ptr [[TMP47]], align 8
// CHECK17-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK17-NEXT:    store ptr @.offload_maptypes, ptr [[TMP48]], align 8
// CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK17-NEXT:    store ptr null, ptr [[TMP49]], align 8
// CHECK17-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK17-NEXT:    store ptr null, ptr [[TMP50]], align 8
// CHECK17-NEXT:    [[TMP51:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK17-NEXT:    store i64 0, ptr [[TMP51]], align 8
// CHECK17-NEXT:    [[TMP52:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK17-NEXT:    store i64 0, ptr [[TMP52]], align 8
// CHECK17-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK17-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP53]], align 4
// CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK17-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP54]], align 4
// CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK17-NEXT:    store i32 0, ptr [[TMP55]], align 4
// CHECK17-NEXT:    [[TMP56:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPg_l152.region_id, ptr [[KERNEL_ARGS]])
// CHECK17-NEXT:    [[TMP57:%.*]] = icmp ne i32 [[TMP56]], 0
// CHECK17-NEXT:    br i1 [[TMP57]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK17:       omp_offload.failed:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPg_l152(ptr [[TMP8]], ptr [[TMP9]], i64 [[TMP1]], ptr [[TMP10]], i64 [[TMP3]], i64 [[TMP5]], ptr [[VLA]], i64 [[TMP12]]) #[[ATTR4:[0-9]+]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[TMP58:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
// CHECK17-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP58]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPg_l152
// CHECK17-SAME: (ptr [[A:%.*]], ptr [[S:%.*]], i64 [[VLA:%.*]], ptr [[VLA1:%.*]], i64 [[VLA2:%.*]], i64 [[VLA4:%.*]], ptr nonnull align 8 dereferenceable(8) [[VLA26:%.*]], i64 [[N:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[VLA_ADDR3:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR5:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA2_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA2]], ptr [[VLA_ADDR3]], align 8
// CHECK17-NEXT:    store i64 [[VLA4]], ptr [[VLA_ADDR5]], align 8
// CHECK17-NEXT:    store ptr [[VLA26]], ptr [[VLA2_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR3]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR5]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VLA2_ADDR]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 8, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPg_l152.omp_outlined, ptr [[TMP4]], ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP5]], ptr [[TMP6]], i64 [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPg_l152.omp_outlined
// CHECK17-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr [[S:%.*]], ptr nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], ptr [[VLA1:%.*]], ptr [[A:%.*]], i64 [[VLA2:%.*]], i64 [[VLA4:%.*]], ptr nonnull align 8 dereferenceable(8) [[VLA26:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[VLA_ADDR3:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR5:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA2_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA2]], ptr [[VLA_ADDR3]], align 8
// CHECK17-NEXT:    store i64 [[VLA4]], ptr [[VLA_ADDR5]], align 8
// CHECK17-NEXT:    store ptr [[VLA26]], ptr [[VLA2_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR3]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load i64, ptr [[VLA_ADDR5]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[VLA2_ADDR]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = call ptr @llvm.stacksave.p0()
// CHECK17-NEXT:    store ptr [[TMP5]], ptr [[SAVED_STACK]], align 8
// CHECK17-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP2]], [[TMP3]]
// CHECK17-NEXT:    [[VLA7:%.*]] = alloca double, i64 [[TMP6]], align 128
// CHECK17-NEXT:    store i64 [[TMP2]], ptr [[__VLA_EXPR0]], align 8
// CHECK17-NEXT:    store i64 [[TMP3]], ptr [[__VLA_EXPR1]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = mul nuw i64 [[TMP2]], [[TMP3]]
// CHECK17-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 8
// CHECK17-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[VLA7]], ptr align 128 [[TMP4]], i64 [[TMP8]], i1 false)
// CHECK17-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[TMP9]], i64 0
// CHECK17-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    call void @_ZN2St7St_funcEPS_iPg(ptr nonnull align 4 dereferenceable(8) [[ARRAYIDX]], ptr [[TMP10]], i32 signext [[TMP11]], ptr [[TMP12]])
// CHECK17-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
// CHECK17-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP13]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@_ZN2St7St_funcEPS_iPg
// CHECK17-SAME: (ptr nonnull align 4 dereferenceable(8) [[THIS:%.*]], ptr [[S:%.*]], i32 signext [[N:%.*]], ptr [[VLA1:%.*]]) #[[ATTR0]] comdat {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [10 x ptr], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [10 x ptr], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [10 x ptr], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 8
// CHECK17-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
// CHECK17-NEXT:    [[TMP6:%.*]] = call ptr @llvm.stacksave.p0()
// CHECK17-NEXT:    store ptr [[TMP6]], ptr [[SAVED_STACK]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = mul nuw i64 [[TMP3]], [[TMP5]]
// CHECK17-NEXT:    [[VLA:%.*]] = alloca double, i64 [[TMP7]], align 128
// CHECK17-NEXT:    store i64 [[TMP3]], ptr [[__VLA_EXPR0]], align 8
// CHECK17-NEXT:    store i64 [[TMP5]], ptr [[__VLA_EXPR1]], align 8
// CHECK17-NEXT:    [[B:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[THIS1]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, ptr [[B]], align 4
// CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 0
// CHECK17-NEXT:    store i32 [[TMP8]], ptr [[A]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP11]], ptr [[N_CASTED]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK17-NEXT:    [[TMP13:%.*]] = mul nuw i64 [[TMP3]], [[TMP5]]
// CHECK17-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 8
// CHECK17-NEXT:    [[B2:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 1
// CHECK17-NEXT:    [[A3:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[B2]], i32 1
// CHECK17-NEXT:    [[TMP16:%.*]] = ptrtoint ptr [[TMP15]] to i64
// CHECK17-NEXT:    [[TMP17:%.*]] = ptrtoint ptr [[A3]] to i64
// CHECK17-NEXT:    [[TMP18:%.*]] = sub i64 [[TMP16]], [[TMP17]]
// CHECK17-NEXT:    [[TMP19:%.*]] = sdiv exact i64 [[TMP18]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
// CHECK17-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes.1, i64 80, i1 false)
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    store ptr [[TMP9]], ptr [[TMP20]], align 8
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    store ptr [[TMP9]], ptr [[TMP21]], align 8
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    store i64 [[TMP1]], ptr [[TMP23]], align 8
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    store i64 [[TMP1]], ptr [[TMP24]], align 8
// CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK17-NEXT:    store ptr null, ptr [[TMP25]], align 8
// CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    store ptr [[TMP10]], ptr [[TMP26]], align 8
// CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    store ptr [[TMP10]], ptr [[TMP27]], align 8
// CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK17-NEXT:    store ptr null, ptr [[TMP28]], align 8
// CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP3]], ptr [[TMP29]], align 8
// CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP3]], ptr [[TMP30]], align 8
// CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK17-NEXT:    store ptr null, ptr [[TMP31]], align 8
// CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK17-NEXT:    store i64 [[TMP5]], ptr [[TMP32]], align 8
// CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK17-NEXT:    store i64 [[TMP5]], ptr [[TMP33]], align 8
// CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK17-NEXT:    store ptr null, ptr [[TMP34]], align 8
// CHECK17-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK17-NEXT:    store ptr [[VLA]], ptr [[TMP35]], align 8
// CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK17-NEXT:    store ptr [[VLA]], ptr [[TMP36]], align 8
// CHECK17-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [10 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK17-NEXT:    store i64 [[TMP14]], ptr [[TMP37]], align 8
// CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
// CHECK17-NEXT:    store ptr null, ptr [[TMP38]], align 8
// CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK17-NEXT:    store ptr [[THIS1]], ptr [[TMP39]], align 8
// CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK17-NEXT:    store ptr [[A3]], ptr [[TMP40]], align 8
// CHECK17-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [10 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK17-NEXT:    store i64 [[TMP19]], ptr [[TMP41]], align 8
// CHECK17-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 6
// CHECK17-NEXT:    store ptr null, ptr [[TMP42]], align 8
// CHECK17-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK17-NEXT:    store ptr [[THIS1]], ptr [[TMP43]], align 8
// CHECK17-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK17-NEXT:    store ptr [[B2]], ptr [[TMP44]], align 8
// CHECK17-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 7
// CHECK17-NEXT:    store ptr null, ptr [[TMP45]], align 8
// CHECK17-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK17-NEXT:    store ptr [[THIS1]], ptr [[TMP46]], align 8
// CHECK17-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK17-NEXT:    store ptr [[A3]], ptr [[TMP47]], align 8
// CHECK17-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 8
// CHECK17-NEXT:    store ptr null, ptr [[TMP48]], align 8
// CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 9
// CHECK17-NEXT:    store i64 [[TMP12]], ptr [[TMP49]], align 8
// CHECK17-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 9
// CHECK17-NEXT:    store i64 [[TMP12]], ptr [[TMP50]], align 8
// CHECK17-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 9
// CHECK17-NEXT:    store ptr null, ptr [[TMP51]], align 8
// CHECK17-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [10 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK17-NEXT:    store i32 3, ptr [[TMP55]], align 4
// CHECK17-NEXT:    [[TMP56:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK17-NEXT:    store i32 10, ptr [[TMP56]], align 4
// CHECK17-NEXT:    [[TMP57:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK17-NEXT:    store ptr [[TMP52]], ptr [[TMP57]], align 8
// CHECK17-NEXT:    [[TMP58:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK17-NEXT:    store ptr [[TMP53]], ptr [[TMP58]], align 8
// CHECK17-NEXT:    [[TMP59:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK17-NEXT:    store ptr [[TMP54]], ptr [[TMP59]], align 8
// CHECK17-NEXT:    [[TMP60:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK17-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP60]], align 8
// CHECK17-NEXT:    [[TMP61:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK17-NEXT:    store ptr null, ptr [[TMP61]], align 8
// CHECK17-NEXT:    [[TMP62:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK17-NEXT:    store ptr null, ptr [[TMP62]], align 8
// CHECK17-NEXT:    [[TMP63:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK17-NEXT:    store i64 0, ptr [[TMP63]], align 8
// CHECK17-NEXT:    [[TMP64:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK17-NEXT:    store i64 0, ptr [[TMP64]], align 8
// CHECK17-NEXT:    [[TMP65:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK17-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP65]], align 4
// CHECK17-NEXT:    [[TMP66:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK17-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP66]], align 4
// CHECK17-NEXT:    [[TMP67:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK17-NEXT:    store i32 0, ptr [[TMP67]], align 4
// CHECK17-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPg_l144.region_id, ptr [[KERNEL_ARGS]])
// CHECK17-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
// CHECK17-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK17:       omp_offload.failed:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPg_l144(ptr [[TMP9]], i64 [[TMP1]], ptr [[TMP10]], i64 [[TMP3]], i64 [[TMP5]], ptr [[VLA]], ptr [[THIS1]], i64 [[TMP12]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[TMP70:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
// CHECK17-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP70]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPg_l144
// CHECK17-SAME: (ptr [[S:%.*]], i64 [[VLA:%.*]], ptr [[VLA1:%.*]], i64 [[VLA2:%.*]], i64 [[VLA4:%.*]], ptr nonnull align 8 dereferenceable(8) [[VLA26:%.*]], ptr [[THIS:%.*]], i64 [[N:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[VLA_ADDR3:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR5:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA2_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA2]], ptr [[VLA_ADDR3]], align 8
// CHECK17-NEXT:    store i64 [[VLA4]], ptr [[VLA_ADDR5]], align 8
// CHECK17-NEXT:    store ptr [[VLA26]], ptr [[VLA2_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR3]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR5]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VLA2_ADDR]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 8, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPg_l144.omp_outlined, i64 [[TMP0]], ptr [[TMP5]], ptr [[TMP4]], i64 [[TMP1]], i64 [[TMP2]], ptr [[TMP3]], ptr [[N_ADDR]], ptr [[TMP6]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPg_l144.omp_outlined
// CHECK17-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[VLA:%.*]], ptr [[VLA1:%.*]], ptr [[THIS:%.*]], i64 [[VLA2:%.*]], i64 [[VLA4:%.*]], ptr nonnull align 8 dereferenceable(8) [[VLA26:%.*]], ptr nonnull align 4 dereferenceable(4) [[N:%.*]], ptr [[S:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[VLA_ADDR3:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR5:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA2_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA2]], ptr [[VLA_ADDR3]], align 8
// CHECK17-NEXT:    store i64 [[VLA4]], ptr [[VLA_ADDR5]], align 8
// CHECK17-NEXT:    store ptr [[VLA26]], ptr [[VLA2_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
// CHECK17-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR3]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load i64, ptr [[VLA_ADDR5]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[VLA2_ADDR]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP6:%.*]] = call ptr @llvm.stacksave.p0()
// CHECK17-NEXT:    store ptr [[TMP6]], ptr [[SAVED_STACK]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = mul nuw i64 [[TMP2]], [[TMP3]]
// CHECK17-NEXT:    [[VLA7:%.*]] = alloca double, i64 [[TMP7]], align 128
// CHECK17-NEXT:    store i64 [[TMP2]], ptr [[__VLA_EXPR0]], align 8
// CHECK17-NEXT:    store i64 [[TMP3]], ptr [[__VLA_EXPR1]], align 8
// CHECK17-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP2]], [[TMP3]]
// CHECK17-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
// CHECK17-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[VLA7]], ptr align 128 [[TMP4]], i64 [[TMP9]], i1 false)
// CHECK17-NEXT:    [[B:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[TMP1]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, ptr [[B]], align 4
// CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[TMP1]], i32 0, i32 0
// CHECK17-NEXT:    store i32 [[TMP10]], ptr [[A]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP10]] to double
// CHECK17-NEXT:    [[TMP11:%.*]] = mul nsw i64 1, [[TMP3]]
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[VLA7]], i64 [[TMP11]]
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[SUB]] to i64
// CHECK17-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds double, ptr [[ARRAYIDX]], i64 [[IDXPROM]]
// CHECK17-NEXT:    store double [[CONV]], ptr [[ARRAYIDX8]], align 8
// CHECK17-NEXT:    [[CONV9:%.*]] = fpext double [[CONV]] to ppc_fp128
// CHECK17-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 8
// CHECK17-NEXT:    [[B10:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[TMP1]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, ptr [[B10]], align 4
// CHECK17-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP14]] to i64
// CHECK17-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds ppc_fp128, ptr [[TMP13]], i64 [[IDXPROM11]]
// CHECK17-NEXT:    store ppc_fp128 [[CONV9]], ptr [[ARRAYIDX12]], align 16
// CHECK17-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
// CHECK17-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP15]])
// CHECK17-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@_Z10array_funcPfP2StiPe
// CHECK19-SAME: (ptr [[A:%.*]], ptr [[S:%.*]], i32 [[N:%.*]], ptr [[VLA1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [8 x ptr], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [8 x ptr], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [8 x ptr], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [8 x i64], align 4
// CHECK19-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK19-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = call ptr @llvm.stacksave.p0()
// CHECK19-NEXT:    store ptr [[TMP3]], ptr [[SAVED_STACK]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP1]], [[TMP2]]
// CHECK19-NEXT:    [[VLA:%.*]] = alloca double, i32 [[TMP4]], align 128
// CHECK19-NEXT:    store i32 [[TMP1]], ptr [[__VLA_EXPR0]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[__VLA_EXPR1]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], ptr [[N_CASTED]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = mul nuw i32 [[TMP1]], [[TMP2]]
// CHECK19-NEXT:    [[TMP11:%.*]] = mul nuw i32 [[TMP10]], 8
// CHECK19-NEXT:    [[TMP12:%.*]] = sext i32 [[TMP11]] to i64
// CHECK19-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes, i32 64, i1 false)
// CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    store ptr [[TMP5]], ptr [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    store ptr [[TMP5]], ptr [[TMP14]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store ptr null, ptr [[TMP15]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK19-NEXT:    store ptr [[TMP6]], ptr [[TMP16]], align 4
// CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK19-NEXT:    store ptr [[TMP6]], ptr [[TMP17]], align 4
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store ptr null, ptr [[TMP18]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    store i32 [[TMP0]], ptr [[TMP19]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    store i32 [[TMP0]], ptr [[TMP20]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store ptr null, ptr [[TMP21]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK19-NEXT:    store ptr [[TMP7]], ptr [[TMP22]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK19-NEXT:    store ptr [[TMP7]], ptr [[TMP23]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK19-NEXT:    store ptr null, ptr [[TMP24]], align 4
// CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK19-NEXT:    store i32 [[TMP1]], ptr [[TMP25]], align 4
// CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK19-NEXT:    store i32 [[TMP1]], ptr [[TMP26]], align 4
// CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK19-NEXT:    store ptr null, ptr [[TMP27]], align 4
// CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[TMP28]], align 4
// CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[TMP29]], align 4
// CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
// CHECK19-NEXT:    store ptr null, ptr [[TMP30]], align 4
// CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK19-NEXT:    store ptr [[VLA]], ptr [[TMP31]], align 4
// CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK19-NEXT:    store ptr [[VLA]], ptr [[TMP32]], align 4
// CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [8 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK19-NEXT:    store i64 [[TMP12]], ptr [[TMP33]], align 4
// CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 6
// CHECK19-NEXT:    store ptr null, ptr [[TMP34]], align 4
// CHECK19-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK19-NEXT:    store i32 [[TMP9]], ptr [[TMP35]], align 4
// CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK19-NEXT:    store i32 [[TMP9]], ptr [[TMP36]], align 4
// CHECK19-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 7
// CHECK19-NEXT:    store ptr null, ptr [[TMP37]], align 4
// CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [8 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [8 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK19-NEXT:    store i32 3, ptr [[TMP41]], align 4
// CHECK19-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK19-NEXT:    store i32 8, ptr [[TMP42]], align 4
// CHECK19-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK19-NEXT:    store ptr [[TMP38]], ptr [[TMP43]], align 4
// CHECK19-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK19-NEXT:    store ptr [[TMP39]], ptr [[TMP44]], align 4
// CHECK19-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK19-NEXT:    store ptr [[TMP40]], ptr [[TMP45]], align 4
// CHECK19-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK19-NEXT:    store ptr @.offload_maptypes, ptr [[TMP46]], align 4
// CHECK19-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK19-NEXT:    store ptr null, ptr [[TMP47]], align 4
// CHECK19-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK19-NEXT:    store ptr null, ptr [[TMP48]], align 4
// CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK19-NEXT:    store i64 0, ptr [[TMP49]], align 8
// CHECK19-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK19-NEXT:    store i64 0, ptr [[TMP50]], align 8
// CHECK19-NEXT:    [[TMP51:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK19-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP51]], align 4
// CHECK19-NEXT:    [[TMP52:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK19-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP52]], align 4
// CHECK19-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK19-NEXT:    store i32 0, ptr [[TMP53]], align 4
// CHECK19-NEXT:    [[TMP54:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPe_l152.region_id, ptr [[KERNEL_ARGS]])
// CHECK19-NEXT:    [[TMP55:%.*]] = icmp ne i32 [[TMP54]], 0
// CHECK19-NEXT:    br i1 [[TMP55]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK19:       omp_offload.failed:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPe_l152(ptr [[TMP5]], ptr [[TMP6]], i32 [[TMP0]], ptr [[TMP7]], i32 [[TMP1]], i32 [[TMP2]], ptr [[VLA]], i32 [[TMP9]]) #[[ATTR4:[0-9]+]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[TMP56:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
// CHECK19-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP56]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPe_l152
// CHECK19-SAME: (ptr [[A:%.*]], ptr [[S:%.*]], i32 [[VLA:%.*]], ptr [[VLA1:%.*]], i32 [[VLA2:%.*]], i32 [[VLA4:%.*]], ptr nonnull align 4 dereferenceable(8) [[VLA26:%.*]], i32 [[N:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[VLA_ADDR3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR5:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA2_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA2]], ptr [[VLA_ADDR3]], align 4
// CHECK19-NEXT:    store i32 [[VLA4]], ptr [[VLA_ADDR5]], align 4
// CHECK19-NEXT:    store ptr [[VLA26]], ptr [[VLA2_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR3]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[VLA_ADDR5]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VLA2_ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK19-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 8, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPe_l152.omp_outlined, ptr [[TMP4]], ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP5]], ptr [[TMP6]], i32 [[TMP1]], i32 [[TMP2]], ptr [[TMP3]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z10array_funcPfP2StiPe_l152.omp_outlined
// CHECK19-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr [[S:%.*]], ptr nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], ptr [[VLA1:%.*]], ptr [[A:%.*]], i32 [[VLA2:%.*]], i32 [[VLA4:%.*]], ptr nonnull align 4 dereferenceable(8) [[VLA26:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[VLA_ADDR3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR5:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA2_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA2]], ptr [[VLA_ADDR3]], align 4
// CHECK19-NEXT:    store i32 [[VLA4]], ptr [[VLA_ADDR5]], align 4
// CHECK19-NEXT:    store ptr [[VLA26]], ptr [[VLA2_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[VLA_ADDR3]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, ptr [[VLA_ADDR5]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[VLA2_ADDR]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = call ptr @llvm.stacksave.p0()
// CHECK19-NEXT:    store ptr [[TMP5]], ptr [[SAVED_STACK]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = mul nuw i32 [[TMP2]], [[TMP3]]
// CHECK19-NEXT:    [[VLA7:%.*]] = alloca double, i32 [[TMP6]], align 128
// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[__VLA_EXPR0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], ptr [[__VLA_EXPR1]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = mul nuw i32 [[TMP2]], [[TMP3]]
// CHECK19-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 8
// CHECK19-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 128 [[VLA7]], ptr align 128 [[TMP4]], i32 [[TMP8]], i1 false)
// CHECK19-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[TMP9]], i32 0
// CHECK19-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    call void @_ZN2St7St_funcEPS_iPe(ptr nonnull align 4 dereferenceable(8) [[ARRAYIDX]], ptr [[TMP10]], i32 [[TMP11]], ptr [[TMP12]])
// CHECK19-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
// CHECK19-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP13]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@_ZN2St7St_funcEPS_iPe
// CHECK19-SAME: (ptr nonnull align 4 dereferenceable(8) [[THIS:%.*]], ptr [[S:%.*]], i32 [[N:%.*]], ptr [[VLA1:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [10 x ptr], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [10 x ptr], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [10 x ptr], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 4
// CHECK19-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = call ptr @llvm.stacksave.p0()
// CHECK19-NEXT:    store ptr [[TMP3]], ptr [[SAVED_STACK]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP1]], [[TMP2]]
// CHECK19-NEXT:    [[VLA:%.*]] = alloca double, i32 [[TMP4]], align 128
// CHECK19-NEXT:    store i32 [[TMP1]], ptr [[__VLA_EXPR0]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[__VLA_EXPR1]], align 4
// CHECK19-NEXT:    [[B:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[THIS1]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, ptr [[B]], align 4
// CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 0
// CHECK19-NEXT:    store i32 [[TMP5]], ptr [[A]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], ptr [[N_CASTED]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = mul nuw i32 [[TMP1]], [[TMP2]]
// CHECK19-NEXT:    [[TMP11:%.*]] = mul nuw i32 [[TMP10]], 8
// CHECK19-NEXT:    [[TMP12:%.*]] = sext i32 [[TMP11]] to i64
// CHECK19-NEXT:    [[B2:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 1
// CHECK19-NEXT:    [[A3:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[B2]], i32 1
// CHECK19-NEXT:    [[TMP14:%.*]] = ptrtoint ptr [[TMP13]] to i64
// CHECK19-NEXT:    [[TMP15:%.*]] = ptrtoint ptr [[A3]] to i64
// CHECK19-NEXT:    [[TMP16:%.*]] = sub i64 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
// CHECK19-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes.1, i32 80, i1 false)
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    store ptr [[TMP6]], ptr [[TMP18]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    store ptr [[TMP6]], ptr [[TMP19]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store ptr null, ptr [[TMP20]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK19-NEXT:    store i32 [[TMP0]], ptr [[TMP21]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK19-NEXT:    store i32 [[TMP0]], ptr [[TMP22]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store ptr null, ptr [[TMP23]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    store ptr [[TMP7]], ptr [[TMP24]], align 4
// CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    store ptr [[TMP7]], ptr [[TMP25]], align 4
// CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store ptr null, ptr [[TMP26]], align 4
// CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK19-NEXT:    store i32 [[TMP1]], ptr [[TMP27]], align 4
// CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK19-NEXT:    store i32 [[TMP1]], ptr [[TMP28]], align 4
// CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK19-NEXT:    store ptr null, ptr [[TMP29]], align 4
// CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[TMP30]], align 4
// CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[TMP31]], align 4
// CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK19-NEXT:    store ptr null, ptr [[TMP32]], align 4
// CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK19-NEXT:    store ptr [[VLA]], ptr [[TMP33]], align 4
// CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK19-NEXT:    store ptr [[VLA]], ptr [[TMP34]], align 4
// CHECK19-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [10 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK19-NEXT:    store i64 [[TMP12]], ptr [[TMP35]], align 4
// CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
// CHECK19-NEXT:    store ptr null, ptr [[TMP36]], align 4
// CHECK19-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK19-NEXT:    store ptr [[THIS1]], ptr [[TMP37]], align 4
// CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK19-NEXT:    store ptr [[A3]], ptr [[TMP38]], align 4
// CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [10 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK19-NEXT:    store i64 [[TMP17]], ptr [[TMP39]], align 4
// CHECK19-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 6
// CHECK19-NEXT:    store ptr null, ptr [[TMP40]], align 4
// CHECK19-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK19-NEXT:    store ptr [[THIS1]], ptr [[TMP41]], align 4
// CHECK19-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK19-NEXT:    store ptr [[B2]], ptr [[TMP42]], align 4
// CHECK19-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 7
// CHECK19-NEXT:    store ptr null, ptr [[TMP43]], align 4
// CHECK19-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK19-NEXT:    store ptr [[THIS1]], ptr [[TMP44]], align 4
// CHECK19-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK19-NEXT:    store ptr [[A3]], ptr [[TMP45]], align 4
// CHECK19-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 8
// CHECK19-NEXT:    store ptr null, ptr [[TMP46]], align 4
// CHECK19-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 9
// CHECK19-NEXT:    store i32 [[TMP9]], ptr [[TMP47]], align 4
// CHECK19-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 9
// CHECK19-NEXT:    store i32 [[TMP9]], ptr [[TMP48]], align 4
// CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 9
// CHECK19-NEXT:    store ptr null, ptr [[TMP49]], align 4
// CHECK19-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [10 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [10 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK19-NEXT:    store i32 3, ptr [[TMP53]], align 4
// CHECK19-NEXT:    [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK19-NEXT:    store i32 10, ptr [[TMP54]], align 4
// CHECK19-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK19-NEXT:    store ptr [[TMP50]], ptr [[TMP55]], align 4
// CHECK19-NEXT:    [[TMP56:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK19-NEXT:    store ptr [[TMP51]], ptr [[TMP56]], align 4
// CHECK19-NEXT:    [[TMP57:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK19-NEXT:    store ptr [[TMP52]], ptr [[TMP57]], align 4
// CHECK19-NEXT:    [[TMP58:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK19-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP58]], align 4
// CHECK19-NEXT:    [[TMP59:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK19-NEXT:    store ptr null, ptr [[TMP59]], align 4
// CHECK19-NEXT:    [[TMP60:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK19-NEXT:    store ptr null, ptr [[TMP60]], align 4
// CHECK19-NEXT:    [[TMP61:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK19-NEXT:    store i64 0, ptr [[TMP61]], align 8
// CHECK19-NEXT:    [[TMP62:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK19-NEXT:    store i64 0, ptr [[TMP62]], align 8
// CHECK19-NEXT:    [[TMP63:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK19-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP63]], align 4
// CHECK19-NEXT:    [[TMP64:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK19-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP64]], align 4
// CHECK19-NEXT:    [[TMP65:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK19-NEXT:    store i32 0, ptr [[TMP65]], align 4
// CHECK19-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPe_l144.region_id, ptr [[KERNEL_ARGS]])
// CHECK19-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
// CHECK19-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK19:       omp_offload.failed:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPe_l144(ptr [[TMP6]], i32 [[TMP0]], ptr [[TMP7]], i32 [[TMP1]], i32 [[TMP2]], ptr [[VLA]], ptr [[THIS1]], i32 [[TMP9]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[TMP68:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
// CHECK19-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP68]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPe_l144
// CHECK19-SAME: (ptr [[S:%.*]], i32 [[VLA:%.*]], ptr [[VLA1:%.*]], i32 [[VLA2:%.*]], i32 [[VLA4:%.*]], ptr nonnull align 4 dereferenceable(8) [[VLA26:%.*]], ptr [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[VLA_ADDR3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR5:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA2_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA2]], ptr [[VLA_ADDR3]], align 4
// CHECK19-NEXT:    store i32 [[VLA4]], ptr [[VLA_ADDR5]], align 4
// CHECK19-NEXT:    store ptr [[VLA26]], ptr [[VLA2_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR3]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[VLA_ADDR5]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VLA2_ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 8, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPe_l144.omp_outlined, i32 [[TMP0]], ptr [[TMP5]], ptr [[TMP4]], i32 [[TMP1]], i32 [[TMP2]], ptr [[TMP3]], ptr [[N_ADDR]], ptr [[TMP6]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2St7St_funcEPS_iPe_l144.omp_outlined
// CHECK19-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i32 [[VLA:%.*]], ptr [[VLA1:%.*]], ptr [[THIS:%.*]], i32 [[VLA2:%.*]], i32 [[VLA4:%.*]], ptr nonnull align 4 dereferenceable(8) [[VLA26:%.*]], ptr nonnull align 4 dereferenceable(4) [[N:%.*]], ptr [[S:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA1_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[VLA_ADDR3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR5:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA2_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
// CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[VLA1]], ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA2]], ptr [[VLA_ADDR3]], align 4
// CHECK19-NEXT:    store i32 [[VLA4]], ptr [[VLA_ADDR5]], align 4
// CHECK19-NEXT:    store ptr [[VLA26]], ptr [[VLA2_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[VLA_ADDR3]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, ptr [[VLA_ADDR5]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[VLA2_ADDR]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = call ptr @llvm.stacksave.p0()
// CHECK19-NEXT:    store ptr [[TMP6]], ptr [[SAVED_STACK]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = mul nuw i32 [[TMP2]], [[TMP3]]
// CHECK19-NEXT:    [[VLA7:%.*]] = alloca double, i32 [[TMP7]], align 128
// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[__VLA_EXPR0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], ptr [[__VLA_EXPR1]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP2]], [[TMP3]]
// CHECK19-NEXT:    [[TMP9:%.*]] = mul nuw i32 [[TMP8]], 8
// CHECK19-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 128 [[VLA7]], ptr align 128 [[TMP4]], i32 [[TMP9]], i1 false)
// CHECK19-NEXT:    [[B:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[TMP1]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, ptr [[B]], align 4
// CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[TMP1]], i32 0, i32 0
// CHECK19-NEXT:    store i32 [[TMP10]], ptr [[A]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP10]] to double
// CHECK19-NEXT:    [[TMP11:%.*]] = mul nsw i32 1, [[TMP3]]
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[VLA7]], i32 [[TMP11]]
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds double, ptr [[ARRAYIDX]], i32 [[SUB]]
// CHECK19-NEXT:    store double [[CONV]], ptr [[ARRAYIDX8]], align 8
// CHECK19-NEXT:    [[CONV9:%.*]] = fpext double [[CONV]] to x86_fp80
// CHECK19-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[VLA1_ADDR]], align 4
// CHECK19-NEXT:    [[B10:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[TMP1]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, ptr [[B10]], align 4
// CHECK19-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds x86_fp80, ptr [[TMP13]], i32 [[TMP14]]
// CHECK19-NEXT:    store x86_fp80 [[CONV9]], ptr [[ARRAYIDX11]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
// CHECK19-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP15]])
// CHECK19-NEXT:    ret void
//