// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
// RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK2
// RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK4

// RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"

// RUN: %clang_cc1  -verify -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=CHECK8
// RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1  -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK9
// RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK10
// RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK11

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

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

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

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

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

    #pragma omp target
    #pragma omp teams
    #pragma omp distribute parallel for firstprivate(g, g1, svar, sfvar)
    for (int i = 0; i < 2; ++i) {

      // addr alloca's

      // private alloca's

      // transfer input parameters into addr alloca's


      // init private alloca's with addr alloca's
      // g

      // g1

      // svar

      // sfvar

      // pass firstprivate parameters to parallel outlined function
      // g

      // g1

      // svar

      // sfvar


      // skip initial params

      // addr alloca's

      // private alloca's (only for 32-bit)

      // transfer input parameters into addr alloca's

      // prepare parameters for lambda
      // g

      // g1

      // svar

      // sfvar

      g = 1;
      g1 = 1;
      svar = 3;
      sfvar = 4.0;

      // pass params to inner lambda
      [&]() {
	g = 2;
	g1 = 2;
	svar = 4;
	sfvar = 8.0;

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

  #pragma omp target
  #pragma omp teams
  #pragma omp distribute parallel for firstprivate(t_var, vec, s_arr, s_arr, var, var, svar)
  for (int i = 0; i < 2; ++i) {
    vec[i] = t_var;
    s_arr[i] = var;
  }
  return tmain<int>();
  #endif
}


// addr alloca's

// skip loop alloca's

// private alloca's


// init addr alloca's with input values

// init private alloca's with addr alloca's
// t-var

// vec

// s_arr

// var

// svar

// pass private alloca's to fork
// not dag to distinguish with S_VAR_CAST

// call destructors: var..

// ..and s_arr


// By OpenMP specifications, 'firstprivate' applies to both distribute and parallel for.
// However, the support for 'firstprivate' of 'parallel' is only used when 'parallel'
// is found alone. Therefore we only have one 'firstprivate' support for 'parallel for'
// in combination

// addr alloca's

// skip loop alloca's

// private alloca's


// init addr alloca's with input values

// init private alloca's with addr alloca's
// vec

// s_arr

// var


// call destructors: var..

// ..and s_arr


// template tmain with S_INT_TY


// addr alloca's

// skip loop alloca's

// private alloca's


// init addr alloca's with input values

// init private alloca's with addr alloca's
// t-var

// vec

// s_arr

// var

// pass private alloca's to fork
// not dag to distinguish with S_VAR_CAST

// call destructors: var..

// ..and s_arr


// By OpenMP specifications, 'firstprivate' applies to both distribute and parallel for.
// However, the support for 'firstprivate' of 'parallel' is only used when 'parallel'
// is found alone. Therefore we only have one 'firstprivate' support for 'parallel for'
// in combination

// addr alloca's

// skip loop alloca's

// private alloca's


// init addr alloca's with input values

// init private alloca's with addr alloca's
// vec

// s_arr

// var


// call destructors: var..

// ..and s_arr


#endif
// CHECK1-LABEL: define {{[^@]+}}@main
// CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK1-NEXT:    [[G1:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK1-NEXT:    store double* [[G]], double** [[G1]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 0
// CHECK1-NEXT:    store double* [[G]], double** [[TMP0]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP2:%.*]] = load double*, double** [[G1]], align 8
// CHECK1-NEXT:    store double* [[TMP2]], double** [[TMP1]], align 8
// CHECK1-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 8 dereferenceable(16) [[REF_TMP]])
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67
// CHECK1-SAME: (i64 [[G:%.*]], i64 [[G1:%.*]], i64 [[SVAR:%.*]], i64 [[SFVAR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    store i64 [[G]], i64* [[G_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[G1]], i64* [[G1_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SVAR]], i64* [[SVAR_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SFVAR]], i64* [[SFVAR_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[G_ADDR]] to double*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[G1_ADDR]] to double*
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SVAR_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[SFVAR_ADDR]] to float*
// CHECK1-NEXT:    store double* [[CONV1]], double** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load double*, double** [[TMP]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, double*, double*, i32*, float*)* @.omp_outlined. to void (i32*, i32*, ...)*), double* [[CONV]], double* [[TMP0]], i32* [[CONV2]], float* [[CONV3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[G:%.*]], double* nonnull align 8 dereferenceable(8) [[G1:%.*]], i32* nonnull align 4 dereferenceable(4) [[SVAR:%.*]], float* nonnull align 4 dereferenceable(4) [[SFVAR:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[G1_ADDR:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[SFVAR_ADDR:%.*]] = alloca float*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[G3:%.*]] = alloca double, align 8
// CHECK1-NEXT:    [[G14:%.*]] = alloca double, align 8
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[SVAR6:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[SFVAR7:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[G_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[G1_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SFVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store double* [[G]], double** [[G_ADDR]], align 8
// CHECK1-NEXT:    store double* [[G1]], double** [[G1_ADDR]], align 8
// CHECK1-NEXT:    store i32* [[SVAR]], i32** [[SVAR_ADDR]], align 8
// CHECK1-NEXT:    store float* [[SFVAR]], float** [[SFVAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load double*, double** [[G_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load double*, double** [[G1_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[SVAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load float*, float** [[SFVAR_ADDR]], align 8
// CHECK1-NEXT:    store double* [[TMP1]], double** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load double*, double** [[TMP]], align 8
// CHECK1-NEXT:    store double* [[TMP4]], double** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP0]], align 8
// CHECK1-NEXT:    store double [[TMP5]], double* [[G3]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[_TMP1]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load double, double* [[TMP6]], align 8
// CHECK1-NEXT:    store double [[TMP7]], double* [[G14]], align 8
// CHECK1-NEXT:    store double* [[G14]], double** [[_TMP5]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[SVAR6]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP3]], align 4
// CHECK1-NEXT:    store float [[TMP9]], float* [[SFVAR7]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 1
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
// CHECK1-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK1-NEXT:    [[TMP21:%.*]] = load double, double* [[G3]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[G_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP21]], double* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = load i64, i64* [[G_CASTED]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = load double*, double** [[_TMP5]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load volatile double, double* [[TMP23]], align 8
// CHECK1-NEXT:    [[CONV9:%.*]] = bitcast i64* [[G1_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP24]], double* [[CONV9]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = load i64, i64* [[G1_CASTED]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[SVAR6]], align 4
// CHECK1-NEXT:    [[CONV10:%.*]] = bitcast i64* [[SVAR_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP26]], i32* [[CONV10]], align 4
// CHECK1-NEXT:    [[TMP27:%.*]] = load i64, i64* [[SVAR_CASTED]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = load float, float* [[SFVAR7]], align 4
// CHECK1-NEXT:    [[CONV11:%.*]] = bitcast i64* [[SFVAR_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP28]], float* [[CONV11]], align 4
// CHECK1-NEXT:    [[TMP29:%.*]] = load i64, i64* [[SFVAR_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP25]], i64 [[TMP27]], i64 [[TMP29]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[G:%.*]], i64 [[G1:%.*]], i64 [[SVAR:%.*]], i64 [[SFVAR:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[G]], i64* [[G_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[G1]], i64* [[G1_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SVAR]], i64* [[SVAR_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SFVAR]], i64* [[SFVAR_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[G_ADDR]] to double*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[G1_ADDR]] to double*
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SVAR_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[SFVAR_ADDR]] to float*
// CHECK1-NEXT:    store double* [[CONV1]], double** [[TMP]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV6:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK1-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV6]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    store double 1.000000e+00, double* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = load double*, double** [[TMP]], align 8
// CHECK1-NEXT:    store volatile double 1.000000e+00, double* [[TMP10]], align 8
// CHECK1-NEXT:    store i32 3, i32* [[CONV2]], align 8
// CHECK1-NEXT:    store float 4.000000e+00, float* [[CONV3]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0
// CHECK1-NEXT:    store double* [[CONV]], double** [[TMP11]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP13:%.*]] = load double*, double** [[TMP]], align 8
// CHECK1-NEXT:    store double* [[TMP13]], double** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 2
// CHECK1-NEXT:    store i32* [[CONV2]], i32** [[TMP14]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 3
// CHECK1-NEXT:    store float* [[CONV3]], float** [[TMP15]], align 8
// CHECK1-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* nonnull align 8 dereferenceable(32) [[REF_TMP]])
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@main
// CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK2-NEXT:    [[G1:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK2-NEXT:    store double* [[G]], double** [[G1]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 0
// CHECK2-NEXT:    store double* [[G]], double** [[TMP0]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP2:%.*]] = load double*, double** [[G1]], align 8
// CHECK2-NEXT:    store double* [[TMP2]], double** [[TMP1]], align 8
// CHECK2-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 8 dereferenceable(16) [[REF_TMP]])
// CHECK2-NEXT:    ret i32 0
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67
// CHECK2-SAME: (i64 [[G:%.*]], i64 [[G1:%.*]], i64 [[SVAR:%.*]], i64 [[SFVAR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    store i64 [[G]], i64* [[G_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[G1]], i64* [[G1_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SVAR]], i64* [[SVAR_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SFVAR]], i64* [[SFVAR_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[G_ADDR]] to double*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[G1_ADDR]] to double*
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SVAR_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[SFVAR_ADDR]] to float*
// CHECK2-NEXT:    store double* [[CONV1]], double** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load double*, double** [[TMP]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, double*, double*, i32*, float*)* @.omp_outlined. to void (i32*, i32*, ...)*), double* [[CONV]], double* [[TMP0]], i32* [[CONV2]], float* [[CONV3]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 8 dereferenceable(8) [[G:%.*]], double* nonnull align 8 dereferenceable(8) [[G1:%.*]], i32* nonnull align 4 dereferenceable(4) [[SVAR:%.*]], float* nonnull align 4 dereferenceable(4) [[SFVAR:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[G1_ADDR:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[SFVAR_ADDR:%.*]] = alloca float*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[G3:%.*]] = alloca double, align 8
// CHECK2-NEXT:    [[G14:%.*]] = alloca double, align 8
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[SVAR6:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SFVAR7:%.*]] = alloca float, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[G_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[G1_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SFVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store double* [[G]], double** [[G_ADDR]], align 8
// CHECK2-NEXT:    store double* [[G1]], double** [[G1_ADDR]], align 8
// CHECK2-NEXT:    store i32* [[SVAR]], i32** [[SVAR_ADDR]], align 8
// CHECK2-NEXT:    store float* [[SFVAR]], float** [[SFVAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load double*, double** [[G_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load double*, double** [[G1_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[SVAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load float*, float** [[SFVAR_ADDR]], align 8
// CHECK2-NEXT:    store double* [[TMP1]], double** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load double*, double** [[TMP]], align 8
// CHECK2-NEXT:    store double* [[TMP4]], double** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP0]], align 8
// CHECK2-NEXT:    store double [[TMP5]], double* [[G3]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[_TMP1]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load double, double* [[TMP6]], align 8
// CHECK2-NEXT:    store double [[TMP7]], double* [[G14]], align 8
// CHECK2-NEXT:    store double* [[G14]], double** [[_TMP5]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[SVAR6]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP3]], align 4
// CHECK2-NEXT:    store float [[TMP9]], float* [[SFVAR7]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 1
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
// CHECK2-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK2-NEXT:    [[TMP21:%.*]] = load double, double* [[G3]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[G_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP21]], double* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = load i64, i64* [[G_CASTED]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = load double*, double** [[_TMP5]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load volatile double, double* [[TMP23]], align 8
// CHECK2-NEXT:    [[CONV9:%.*]] = bitcast i64* [[G1_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP24]], double* [[CONV9]], align 8
// CHECK2-NEXT:    [[TMP25:%.*]] = load i64, i64* [[G1_CASTED]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[SVAR6]], align 4
// CHECK2-NEXT:    [[CONV10:%.*]] = bitcast i64* [[SVAR_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP26]], i32* [[CONV10]], align 4
// CHECK2-NEXT:    [[TMP27:%.*]] = load i64, i64* [[SVAR_CASTED]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = load float, float* [[SFVAR7]], align 4
// CHECK2-NEXT:    [[CONV11:%.*]] = bitcast i64* [[SFVAR_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP28]], float* [[CONV11]], align 4
// CHECK2-NEXT:    [[TMP29:%.*]] = load i64, i64* [[SFVAR_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP25]], i64 [[TMP27]], i64 [[TMP29]])
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[G:%.*]], i64 [[G1:%.*]], i64 [[SVAR:%.*]], i64 [[SFVAR:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[G]], i64* [[G_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[G1]], i64* [[G1_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SVAR]], i64* [[SVAR_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SFVAR]], i64* [[SFVAR_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[G_ADDR]] to double*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[G1_ADDR]] to double*
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SVAR_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[SFVAR_ADDR]] to float*
// CHECK2-NEXT:    store double* [[CONV1]], double** [[TMP]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV6:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV6]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    store double 1.000000e+00, double* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP10:%.*]] = load double*, double** [[TMP]], align 8
// CHECK2-NEXT:    store volatile double 1.000000e+00, double* [[TMP10]], align 8
// CHECK2-NEXT:    store i32 3, i32* [[CONV2]], align 8
// CHECK2-NEXT:    store float 4.000000e+00, float* [[CONV3]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0
// CHECK2-NEXT:    store double* [[CONV]], double** [[TMP11]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP13:%.*]] = load double*, double** [[TMP]], align 8
// CHECK2-NEXT:    store double* [[TMP13]], double** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 2
// CHECK2-NEXT:    store i32* [[CONV2]], i32** [[TMP14]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 3
// CHECK2-NEXT:    store float* [[CONV3]], float** [[TMP15]], align 8
// CHECK2-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* nonnull align 8 dereferenceable(32) [[REF_TMP]])
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK2-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK2-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[G1:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK3-NEXT:    store double* [[G]], double** [[G1]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 0
// CHECK3-NEXT:    store double* [[G]], double** [[TMP0]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP2:%.*]] = load double*, double** [[G1]], align 4
// CHECK3-NEXT:    store double* [[TMP2]], double** [[TMP1]], align 4
// CHECK3-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 4 dereferenceable(8) [[REF_TMP]])
// CHECK3-NEXT:    ret i32 0
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67
// CHECK3-SAME: (double* nonnull align 4 dereferenceable(8) [[G:%.*]], double* nonnull align 4 dereferenceable(8) [[G1:%.*]], i32 [[SVAR:%.*]], i32 [[SFVAR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[G1_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[G2:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[G13:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[_TMP4:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    store double* [[G]], double** [[G_ADDR]], align 4
// CHECK3-NEXT:    store double* [[G1]], double** [[G1_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SVAR]], i32* [[SVAR_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SFVAR]], i32* [[SFVAR_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load double*, double** [[G_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** [[G1_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[SFVAR_ADDR]] to float*
// CHECK3-NEXT:    store double* [[TMP1]], double** [[TMP]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load double, double* [[TMP0]], align 8
// CHECK3-NEXT:    store double [[TMP2]], double* [[G2]], align 8
// CHECK3-NEXT:    [[TMP3:%.*]] = load double*, double** [[TMP]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load volatile double, double* [[TMP3]], align 4
// CHECK3-NEXT:    store double [[TMP4]], double* [[G13]], align 8
// CHECK3-NEXT:    store double* [[G13]], double** [[_TMP4]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load double*, double** [[_TMP4]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, double*, double*, i32*, float*)* @.omp_outlined. to void (i32*, i32*, ...)*), double* [[G2]], double* [[TMP5]], i32* [[SVAR_ADDR]], float* [[CONV]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 4 dereferenceable(8) [[G:%.*]], double* nonnull align 4 dereferenceable(8) [[G1:%.*]], i32* nonnull align 4 dereferenceable(4) [[SVAR:%.*]], float* nonnull align 4 dereferenceable(4) [[SFVAR:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[G1_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[SFVAR_ADDR:%.*]] = alloca float*, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[_TMP1:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[G3:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[G14:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[_TMP5:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SVAR6:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SFVAR7:%.*]] = alloca float, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[G1_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SFVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store double* [[G]], double** [[G_ADDR]], align 4
// CHECK3-NEXT:    store double* [[G1]], double** [[G1_ADDR]], align 4
// CHECK3-NEXT:    store i32* [[SVAR]], i32** [[SVAR_ADDR]], align 4
// CHECK3-NEXT:    store float* [[SFVAR]], float** [[SFVAR_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load double*, double** [[G_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** [[G1_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[SVAR_ADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load float*, float** [[SFVAR_ADDR]], align 4
// CHECK3-NEXT:    store double* [[TMP1]], double** [[TMP]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load double*, double** [[TMP]], align 4
// CHECK3-NEXT:    store double* [[TMP4]], double** [[_TMP1]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP0]], align 8
// CHECK3-NEXT:    store double [[TMP5]], double* [[G3]], align 8
// CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[_TMP1]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load double, double* [[TMP6]], align 4
// CHECK3-NEXT:    store double [[TMP7]], double* [[G14]], align 8
// CHECK3-NEXT:    store double* [[G14]], double** [[_TMP5]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[SVAR6]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP3]], align 4
// CHECK3-NEXT:    store float [[TMP9]], float* [[SFVAR7]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 1
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
// CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = load double*, double** [[_TMP5]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = load volatile double, double* [[TMP19]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[G1_CASTED]] to double*
// CHECK3-NEXT:    store double [[TMP20]], double* [[CONV]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[G1_CASTED]], align 4
// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[SVAR6]], align 4
// CHECK3-NEXT:    store i32 [[TMP22]], i32* [[SVAR_CASTED]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[SVAR_CASTED]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = load float, float* [[SFVAR7]], align 4
// CHECK3-NEXT:    [[CONV9:%.*]] = bitcast i32* [[SFVAR_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP24]], float* [[CONV9]], align 4
// CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[SFVAR_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, double*, i32, i32, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], double* [[G3]], i32 [[TMP21]], i32 [[TMP23]], i32 [[TMP25]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], double* nonnull align 4 dereferenceable(8) [[G:%.*]], i32 [[G1:%.*]], i32 [[SVAR:%.*]], i32 [[SFVAR:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[G1_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[G3:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store double* [[G]], double** [[G_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[G1]], i32* [[G1_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SVAR]], i32* [[SVAR_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SFVAR]], i32* [[SFVAR_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load double*, double** [[G_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[G1_ADDR]] to double*
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SFVAR_ADDR]] to float*
// CHECK3-NEXT:    store double* [[CONV]], double** [[TMP]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load double, double* [[TMP0]], align 8
// CHECK3-NEXT:    store double [[TMP3]], double* [[G3]], align 8
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    store double 1.000000e+00, double* [[G3]], align 8
// CHECK3-NEXT:    [[TMP12:%.*]] = load double*, double** [[TMP]], align 4
// CHECK3-NEXT:    store volatile double 1.000000e+00, double* [[TMP12]], align 4
// CHECK3-NEXT:    store i32 3, i32* [[SVAR_ADDR]], align 4
// CHECK3-NEXT:    store float 4.000000e+00, float* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0
// CHECK3-NEXT:    store double* [[G3]], double** [[TMP13]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP15:%.*]] = load double*, double** [[TMP]], align 4
// CHECK3-NEXT:    store double* [[TMP15]], double** [[TMP14]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 2
// CHECK3-NEXT:    store i32* [[SVAR_ADDR]], i32** [[TMP16]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 3
// CHECK3-NEXT:    store float* [[CONV1]], float** [[TMP17]], align 4
// CHECK3-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* nonnull align 4 dereferenceable(16) [[REF_TMP]])
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK3-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@main
// CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[G1:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK4-NEXT:    store double* [[G]], double** [[G1]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 0
// CHECK4-NEXT:    store double* [[G]], double** [[TMP0]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP2:%.*]] = load double*, double** [[G1]], align 4
// CHECK4-NEXT:    store double* [[TMP2]], double** [[TMP1]], align 4
// CHECK4-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 4 dereferenceable(8) [[REF_TMP]])
// CHECK4-NEXT:    ret i32 0
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67
// CHECK4-SAME: (double* nonnull align 4 dereferenceable(8) [[G:%.*]], double* nonnull align 4 dereferenceable(8) [[G1:%.*]], i32 [[SVAR:%.*]], i32 [[SFVAR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[G1_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[G2:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[G13:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[_TMP4:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    store double* [[G]], double** [[G_ADDR]], align 4
// CHECK4-NEXT:    store double* [[G1]], double** [[G1_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SVAR]], i32* [[SVAR_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SFVAR]], i32* [[SFVAR_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load double*, double** [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** [[G1_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[SFVAR_ADDR]] to float*
// CHECK4-NEXT:    store double* [[TMP1]], double** [[TMP]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load double, double* [[TMP0]], align 8
// CHECK4-NEXT:    store double [[TMP2]], double* [[G2]], align 8
// CHECK4-NEXT:    [[TMP3:%.*]] = load double*, double** [[TMP]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load volatile double, double* [[TMP3]], align 4
// CHECK4-NEXT:    store double [[TMP4]], double* [[G13]], align 8
// CHECK4-NEXT:    store double* [[G13]], double** [[_TMP4]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load double*, double** [[_TMP4]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, double*, double*, i32*, float*)* @.omp_outlined. to void (i32*, i32*, ...)*), double* [[G2]], double* [[TMP5]], i32* [[SVAR_ADDR]], float* [[CONV]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], double* nonnull align 4 dereferenceable(8) [[G:%.*]], double* nonnull align 4 dereferenceable(8) [[G1:%.*]], i32* nonnull align 4 dereferenceable(4) [[SVAR:%.*]], float* nonnull align 4 dereferenceable(4) [[SFVAR:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[G1_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[SFVAR_ADDR:%.*]] = alloca float*, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[_TMP1:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[G3:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[G14:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[_TMP5:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SVAR6:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SFVAR7:%.*]] = alloca float, align 4
// CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[G1_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SFVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store double* [[G]], double** [[G_ADDR]], align 4
// CHECK4-NEXT:    store double* [[G1]], double** [[G1_ADDR]], align 4
// CHECK4-NEXT:    store i32* [[SVAR]], i32** [[SVAR_ADDR]], align 4
// CHECK4-NEXT:    store float* [[SFVAR]], float** [[SFVAR_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load double*, double** [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** [[G1_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[SVAR_ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load float*, float** [[SFVAR_ADDR]], align 4
// CHECK4-NEXT:    store double* [[TMP1]], double** [[TMP]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load double*, double** [[TMP]], align 4
// CHECK4-NEXT:    store double* [[TMP4]], double** [[_TMP1]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP0]], align 8
// CHECK4-NEXT:    store double [[TMP5]], double* [[G3]], align 8
// CHECK4-NEXT:    [[TMP6:%.*]] = load double*, double** [[_TMP1]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load double, double* [[TMP6]], align 4
// CHECK4-NEXT:    store double [[TMP7]], double* [[G14]], align 8
// CHECK4-NEXT:    store double* [[G14]], double** [[_TMP5]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK4-NEXT:    store i32 [[TMP8]], i32* [[SVAR6]], align 4
// CHECK4-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP3]], align 4
// CHECK4-NEXT:    store float [[TMP9]], float* [[SFVAR7]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 1
// CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4:       cond.true:
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
// CHECK4-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = load double*, double** [[_TMP5]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = load volatile double, double* [[TMP19]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[G1_CASTED]] to double*
// CHECK4-NEXT:    store double [[TMP20]], double* [[CONV]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[G1_CASTED]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[SVAR6]], align 4
// CHECK4-NEXT:    store i32 [[TMP22]], i32* [[SVAR_CASTED]], align 4
// CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[SVAR_CASTED]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = load float, float* [[SFVAR7]], align 4
// CHECK4-NEXT:    [[CONV9:%.*]] = bitcast i32* [[SFVAR_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP24]], float* [[CONV9]], align 4
// CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[SFVAR_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, double*, i32, i32, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], double* [[G3]], i32 [[TMP21]], i32 [[TMP23]], i32 [[TMP25]])
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK4:       omp.loop.exit:
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], double* nonnull align 4 dereferenceable(8) [[G:%.*]], i32 [[G1:%.*]], i32 [[SVAR:%.*]], i32 [[SFVAR:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[G1_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[G3:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store double* [[G]], double** [[G_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[G1]], i32* [[G1_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SVAR]], i32* [[SVAR_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SFVAR]], i32* [[SFVAR_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load double*, double** [[G_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[G1_ADDR]] to double*
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SFVAR_ADDR]] to float*
// CHECK4-NEXT:    store double* [[CONV]], double** [[TMP]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load double, double* [[TMP0]], align 8
// CHECK4-NEXT:    store double [[TMP3]], double* [[G3]], align 8
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
// CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4:       cond.true:
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK4-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK4-NEXT:    store double 1.000000e+00, double* [[G3]], align 8
// CHECK4-NEXT:    [[TMP12:%.*]] = load double*, double** [[TMP]], align 4
// CHECK4-NEXT:    store volatile double 1.000000e+00, double* [[TMP12]], align 4
// CHECK4-NEXT:    store i32 3, i32* [[SVAR_ADDR]], align 4
// CHECK4-NEXT:    store float 4.000000e+00, float* [[CONV1]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0
// CHECK4-NEXT:    store double* [[G3]], double** [[TMP13]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP15:%.*]] = load double*, double** [[TMP]], align 4
// CHECK4-NEXT:    store double* [[TMP15]], double** [[TMP14]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 2
// CHECK4-NEXT:    store i32* [[SVAR_ADDR]], i32** [[TMP16]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 3
// CHECK4-NEXT:    store float* [[CONV1]], float** [[TMP17]], align 4
// CHECK4-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* nonnull align 4 dereferenceable(16) [[REF_TMP]])
// CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK4:       omp.body.continue:
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK4-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK4:       omp.loop.exit:
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK4-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK4-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@main
// CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK8-NEXT:    [[G1:%.*]] = alloca double*, align 8
// CHECK8-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK8-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK8-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK8-NEXT:    [[VAR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK8-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
// CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK8-NEXT:    store double* [[G]], double** [[G1]], align 8
// CHECK8-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK8-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const.main.vec to i8*), i64 8, i1 false)
// CHECK8-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR]], i64 0, i64 0
// CHECK8-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float 1.000000e+00)
// CHECK8-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_BEGIN]], i64 1
// CHECK8-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float 2.000000e+00)
// CHECK8-NEXT:    store %struct.S* [[TEST]], %struct.S** [[VAR]], align 8
// CHECK8-NEXT:    [[TMP1:%.*]] = load %struct.S*, %struct.S** [[VAR]], align 8
// CHECK8-NEXT:    store %struct.S* [[TMP1]], %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[T_VAR]], align 4
// CHECK8-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32*
// CHECK8-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
// CHECK8-NEXT:    [[TMP3:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8
// CHECK8-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* @_ZZ4mainE4svar, align 4
// CHECK8-NEXT:    [[CONV1:%.*]] = bitcast i64* [[SVAR_CASTED]] to i32*
// CHECK8-NEXT:    store i32 [[TMP5]], i32* [[CONV1]], align 4
// CHECK8-NEXT:    [[TMP6:%.*]] = load i64, i64* [[SVAR_CASTED]], align 8
// CHECK8-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP8:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK8-NEXT:    store i64 [[TMP3]], i64* [[TMP10]], align 8
// CHECK8-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK8-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
// CHECK8-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK8-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK8-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK8-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [2 x i32]**
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP15]], align 8
// CHECK8-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK8-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [2 x i32]**
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP17]], align 8
// CHECK8-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK8-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK8-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK8-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [2 x %struct.S]**
// CHECK8-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[TMP20]], align 8
// CHECK8-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK8-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [2 x %struct.S]**
// CHECK8-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[TMP22]], align 8
// CHECK8-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK8-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK8-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK8-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to %struct.S**
// CHECK8-NEXT:    store %struct.S* [[TMP7]], %struct.S** [[TMP25]], align 8
// CHECK8-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK8-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S**
// CHECK8-NEXT:    store %struct.S* [[TMP8]], %struct.S** [[TMP27]], align 8
// CHECK8-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK8-NEXT:    store i8* null, i8** [[TMP28]], align 8
// CHECK8-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK8-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
// CHECK8-NEXT:    store i64 [[TMP6]], i64* [[TMP30]], align 8
// CHECK8-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK8-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
// CHECK8-NEXT:    store i64 [[TMP6]], i64* [[TMP32]], align 8
// CHECK8-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK8-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK8-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK8-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2)
// CHECK8-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK8-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK8-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK8:       omp_offload.failed:
// CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139(i64 [[TMP3]], [2 x i32]* [[VEC]], [2 x %struct.S]* [[S_ARR]], %struct.S* [[TMP4]], i64 [[TMP6]]) #[[ATTR4:[0-9]+]]
// CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK8:       omp_offload.cont:
// CHECK8-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
// CHECK8-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK8:       arraydestroy.body:
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP38]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK8-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE3:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK8:       arraydestroy.done3:
// CHECK8-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK8-NEXT:    [[TMP39:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK8-NEXT:    ret i32 [[TMP39]]
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK8-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    call void @_ZN1SIfEC2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK8-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK8-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK8-NEXT:    call void @_ZN1SIfEC2Ef(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]], float [[TMP0]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139
// CHECK8-SAME: (i64 [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 [[SVAR:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 8
// CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    store i64 [[SVAR]], i64* [[SVAR_ADDR]], align 8
// CHECK8-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32*
// CHECK8-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    [[CONV1:%.*]] = bitcast i64* [[SVAR_ADDR]] to i32*
// CHECK8-NEXT:    store %struct.S* [[TMP2]], %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP3:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S]*, %struct.S*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], [2 x i32]* [[TMP0]], [2 x %struct.S]* [[TMP1]], %struct.S* [[TMP3]], i32* [[CONV1]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32* nonnull align 4 dereferenceable(4) [[SVAR:%.*]]) #[[ATTR3]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 8
// CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[_TMP1:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK8-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S], align 4
// CHECK8-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK8-NEXT:    [[_TMP8:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[SVAR9:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK8-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    store i32* [[SVAR]], i32** [[SVAR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP3:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[SVAR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S* [[TMP3]], %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP5:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    store %struct.S* [[TMP5]], %struct.S** [[_TMP1]], align 8
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK8-NEXT:    store i32 [[TMP6]], i32* [[T_VAR3]], align 4
// CHECK8-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK8-NEXT:    [[TMP8:%.*]] = bitcast [2 x i32]* [[TMP1]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP7]], i8* align 4 [[TMP8]], i64 8, i1 false)
// CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR5]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP9:%.*]] = bitcast [2 x %struct.S]* [[TMP2]] to %struct.S*
// CHECK8-NEXT:    [[TMP10:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK8-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP10]]
// CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK8:       omp.arraycpy.body:
// CHECK8-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP9]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK8-NEXT:    [[TMP11:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK8-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP11]], i8* align 4 [[TMP12]], i64 4, i1 false)
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK8-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP10]]
// CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK8:       omp.arraycpy.done6:
// CHECK8-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[_TMP1]], align 8
// CHECK8-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[VAR7]] to i8*
// CHECK8-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP14]], i8* align 4 [[TMP15]], i64 4, i1 false)
// CHECK8-NEXT:    store %struct.S* [[VAR7]], %struct.S** [[_TMP8]], align 8
// CHECK8-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK8-NEXT:    store i32 [[TMP16]], i32* [[SVAR9]], align 4
// CHECK8-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP18]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP19]], 1
// CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK8:       cond.true:
// CHECK8-NEXT:    br label [[COND_END:%.*]]
// CHECK8:       cond.false:
// CHECK8-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP20]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP21]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP22]], [[TMP23]]
// CHECK8-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK8:       omp.inner.for.cond.cleanup:
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
// CHECK8-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP27:%.*]] = zext i32 [[TMP26]] to i64
// CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK8-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32*
// CHECK8-NEXT:    store i32 [[TMP28]], i32* [[CONV]], align 4
// CHECK8-NEXT:    [[TMP29:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8
// CHECK8-NEXT:    [[TMP30:%.*]] = load %struct.S*, %struct.S** [[_TMP8]], align 8
// CHECK8-NEXT:    [[TMP31:%.*]] = load i32, i32* [[SVAR9]], align 4
// CHECK8-NEXT:    [[CONV11:%.*]] = bitcast i64* [[SVAR_CASTED]] to i32*
// CHECK8-NEXT:    store i32 [[TMP31]], i32* [[CONV11]], align 4
// CHECK8-NEXT:    [[TMP32:%.*]] = load i64, i64* [[SVAR_CASTED]], align 8
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 7, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [2 x i32]*, i64, [2 x %struct.S]*, %struct.S*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP25]], i64 [[TMP27]], [2 x i32]* [[VEC4]], i64 [[TMP29]], [2 x %struct.S]* [[S_ARR5]], %struct.S* [[TMP30]], i64 [[TMP32]])
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP36]])
// CHECK8-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAY_BEGIN12:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR5]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN12]], i64 2
// CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK8:       arraydestroy.body:
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP37]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK8-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN12]]
// CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE13:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK8:       arraydestroy.done13:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 [[T_VAR:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 [[SVAR:%.*]]) #[[ATTR3]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 8
// CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[VEC5:%.*]] = alloca [2 x i32], align 4
// CHECK8-NEXT:    [[S_ARR6:%.*]] = alloca [2 x %struct.S], align 4
// CHECK8-NEXT:    [[VAR8:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK8-NEXT:    [[_TMP9:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK8-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK8-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8
// CHECK8-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    store i64 [[SVAR]], i64* [[SVAR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32*
// CHECK8-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    [[CONV1:%.*]] = bitcast i64* [[SVAR_ADDR]] to i32*
// CHECK8-NEXT:    store %struct.S* [[TMP2]], %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK8-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP3]] to i32
// CHECK8-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK8-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP4]] to i32
// CHECK8-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = bitcast [2 x i32]* [[VEC5]] to i8*
// CHECK8-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[TMP0]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP5]], i8* align 4 [[TMP6]], i64 8, i1 false)
// CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR6]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP7:%.*]] = bitcast [2 x %struct.S]* [[TMP1]] to %struct.S*
// CHECK8-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK8-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE7:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK8:       omp.arraycpy.body:
// CHECK8-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK8-NEXT:    [[TMP9:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK8-NEXT:    [[TMP10:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP9]], i8* align 4 [[TMP10]], i64 4, i1 false)
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK8-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE7]], label [[OMP_ARRAYCPY_BODY]]
// CHECK8:       omp.arraycpy.done7:
// CHECK8-NEXT:    [[TMP11:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[VAR8]] to i8*
// CHECK8-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP11]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i64 4, i1 false)
// CHECK8-NEXT:    store %struct.S* [[VAR8]], %struct.S** [[_TMP9]], align 8
// CHECK8-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 1
// CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK8:       cond.true:
// CHECK8-NEXT:    br label [[COND_END:%.*]]
// CHECK8:       cond.false:
// CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK8-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK8:       omp.inner.for.cond.cleanup:
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK8-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
// CHECK8-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP23]] to i64
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC5]], i64 0, i64 [[IDXPROM]]
// CHECK8-NEXT:    store i32 [[TMP22]], i32* [[ARRAYIDX]], align 4
// CHECK8-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[_TMP9]], align 8
// CHECK8-NEXT:    [[TMP25:%.*]] = load i32, i32* [[I]], align 4
// CHECK8-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP25]] to i64
// CHECK8-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR6]], i64 0, i64 [[IDXPROM11]]
// CHECK8-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[ARRAYIDX12]] to i8*
// CHECK8-NEXT:    [[TMP27:%.*]] = bitcast %struct.S* [[TMP24]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP26]], i8* align 4 [[TMP27]], i64 4, i1 false)
// CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK8:       omp.body.continue:
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK8-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK8-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR8]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAY_BEGIN14:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR6]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN14]], i64 2
// CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK8:       arraydestroy.body:
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK8-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN14]]
// CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE15:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK8:       arraydestroy.done15:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK8-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    call void @_ZN1SIfED2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK8-SAME: () #[[ATTR5:[0-9]+]] comdat {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK8-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK8-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK8-NEXT:    [[VAR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK8-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
// CHECK8-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK8-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiET_v.vec to i8*), i64 8, i1 false)
// CHECK8-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i64 0, i64 0
// CHECK8-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 signext 1)
// CHECK8-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYINIT_BEGIN]], i64 1
// CHECK8-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 signext 2)
// CHECK8-NEXT:    store %struct.S.0* [[TEST]], %struct.S.0** [[VAR]], align 8
// CHECK8-NEXT:    [[TMP1:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 8
// CHECK8-NEXT:    store %struct.S.0* [[TMP1]], %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[T_VAR]], align 4
// CHECK8-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32*
// CHECK8-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
// CHECK8-NEXT:    [[TMP3:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8
// CHECK8-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP5:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP6:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK8-NEXT:    store i64 [[TMP3]], i64* [[TMP8]], align 8
// CHECK8-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK8-NEXT:    store i64 [[TMP3]], i64* [[TMP10]], align 8
// CHECK8-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK8-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK8-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK8-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to [2 x i32]**
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP13]], align 8
// CHECK8-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK8-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [2 x i32]**
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP15]], align 8
// CHECK8-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK8-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK8-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK8-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [2 x %struct.S.0]**
// CHECK8-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[TMP18]], align 8
// CHECK8-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK8-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [2 x %struct.S.0]**
// CHECK8-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[TMP20]], align 8
// CHECK8-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK8-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK8-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK8-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to %struct.S.0**
// CHECK8-NEXT:    store %struct.S.0* [[TMP5]], %struct.S.0** [[TMP23]], align 8
// CHECK8-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK8-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to %struct.S.0**
// CHECK8-NEXT:    store %struct.S.0* [[TMP6]], %struct.S.0** [[TMP25]], align 8
// CHECK8-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK8-NEXT:    store i8* null, i8** [[TMP26]], align 8
// CHECK8-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK8-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2)
// CHECK8-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK8-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK8-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK8:       omp_offload.failed:
// CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48(i64 [[TMP3]], [2 x i32]* [[VEC]], [2 x %struct.S.0]* [[S_ARR]], %struct.S.0* [[TMP4]]) #[[ATTR4]]
// CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK8:       omp_offload.cont:
// CHECK8-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 2
// CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK8:       arraydestroy.body:
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK8-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK8:       arraydestroy.done2:
// CHECK8-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK8-NEXT:    [[TMP32:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK8-NEXT:    ret i32 [[TMP32]]
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK8-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    store float 0.000000e+00, float* [[F]], align 4
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK8-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK8-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK8-NEXT:    store float [[TMP0]], float* [[F]], align 4
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK8-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK8-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK8-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    call void @_ZN1SIiEC2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK8-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK8-NEXT:    call void @_ZN1SIiEC2Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]], i32 signext [[TMP0]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48
// CHECK8-SAME: (i64 [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32*
// CHECK8-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S.0* [[TMP2]], %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], [2 x i32]* [[TMP0]], [2 x %struct.S.0]* [[TMP1]], %struct.S.0* [[TMP3]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK8-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK8-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK8-NEXT:    [[_TMP8:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK8-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S.0* [[TMP3]], %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    store %struct.S.0* [[TMP4]], %struct.S.0** [[_TMP1]], align 8
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK8-NEXT:    store i32 [[TMP5]], i32* [[T_VAR3]], align 4
// CHECK8-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK8-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[TMP1]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP6]], i8* align 4 [[TMP7]], i64 8, i1 false)
// CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP8:%.*]] = bitcast [2 x %struct.S.0]* [[TMP2]] to %struct.S.0*
// CHECK8-NEXT:    [[TMP9:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 2
// CHECK8-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP9]]
// CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK8:       omp.arraycpy.body:
// CHECK8-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK8-NEXT:    [[TMP10:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK8-NEXT:    [[TMP11:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP10]], i8* align 4 [[TMP11]], i64 4, i1 false)
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK8-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP9]]
// CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK8:       omp.arraycpy.done6:
// CHECK8-NEXT:    [[TMP12:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 8
// CHECK8-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[VAR7]] to i8*
// CHECK8-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[TMP12]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK8-NEXT:    store %struct.S.0* [[VAR7]], %struct.S.0** [[_TMP8]], align 8
// CHECK8-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP16]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP17]], 1
// CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK8:       cond.true:
// CHECK8-NEXT:    br label [[COND_END:%.*]]
// CHECK8:       cond.false:
// CHECK8-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK8-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK8:       omp.inner.for.cond.cleanup:
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP22]] to i64
// CHECK8-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
// CHECK8-NEXT:    [[TMP26:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK8-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32*
// CHECK8-NEXT:    store i32 [[TMP26]], i32* [[CONV]], align 4
// CHECK8-NEXT:    [[TMP27:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8
// CHECK8-NEXT:    [[TMP28:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP8]], align 8
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [2 x i32]*, i64, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP23]], i64 [[TMP25]], [2 x i32]* [[VEC4]], i64 [[TMP27]], [2 x %struct.S.0]* [[S_ARR5]], %struct.S.0* [[TMP28]])
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], [[TMP30]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP32]])
// CHECK8-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN10]], i64 2
// CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK8:       arraydestroy.body:
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP33]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK8-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK8:       arraydestroy.done11:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 [[T_VAR:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK8-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK8-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK8-NEXT:    [[_TMP8:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK8-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK8-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK8-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8
// CHECK8-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK8-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32*
// CHECK8-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK8-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK8-NEXT:    store %struct.S.0* [[TMP2]], %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK8-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP3]] to i32
// CHECK8-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK8-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP4]] to i32
// CHECK8-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK8-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[TMP0]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP5]], i8* align 4 [[TMP6]], i64 8, i1 false)
// CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP7:%.*]] = bitcast [2 x %struct.S.0]* [[TMP1]] to %struct.S.0*
// CHECK8-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 2
// CHECK8-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK8:       omp.arraycpy.body:
// CHECK8-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK8-NEXT:    [[TMP9:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK8-NEXT:    [[TMP10:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP9]], i8* align 4 [[TMP10]], i64 4, i1 false)
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK8-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK8-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK8:       omp.arraycpy.done6:
// CHECK8-NEXT:    [[TMP11:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK8-NEXT:    [[TMP12:%.*]] = bitcast %struct.S.0* [[VAR7]] to i8*
// CHECK8-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[TMP11]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i64 4, i1 false)
// CHECK8-NEXT:    store %struct.S.0* [[VAR7]], %struct.S.0** [[_TMP8]], align 8
// CHECK8-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 1
// CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK8:       cond.true:
// CHECK8-NEXT:    br label [[COND_END:%.*]]
// CHECK8:       cond.false:
// CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK8-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK8:       omp.inner.for.cond.cleanup:
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK8-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
// CHECK8-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP23]] to i64
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC4]], i64 0, i64 [[IDXPROM]]
// CHECK8-NEXT:    store i32 [[TMP22]], i32* [[ARRAYIDX]], align 4
// CHECK8-NEXT:    [[TMP24:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP8]], align 8
// CHECK8-NEXT:    [[TMP25:%.*]] = load i32, i32* [[I]], align 4
// CHECK8-NEXT:    [[IDXPROM10:%.*]] = sext i32 [[TMP25]] to i64
// CHECK8-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i64 0, i64 [[IDXPROM10]]
// CHECK8-NEXT:    [[TMP26:%.*]] = bitcast %struct.S.0* [[ARRAYIDX11]] to i8*
// CHECK8-NEXT:    [[TMP27:%.*]] = bitcast %struct.S.0* [[TMP24]] to i8*
// CHECK8-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP26]], i8* align 4 [[TMP27]], i64 4, i1 false)
// CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK8:       omp.body.continue:
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK8-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK8-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK8-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAY_BEGIN13:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN13]], i64 2
// CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK8:       arraydestroy.body:
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK8-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN13]]
// CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE14:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK8:       arraydestroy.done14:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK8-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    call void @_ZN1SIiED2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK8-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    store i32 0, i32* [[F]], align 4
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK8-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK8-NEXT:    store i32 [[TMP0]], i32* [[F]], align 4
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK8-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK8-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK8-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK8-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK9-NEXT:    [[G1:%.*]] = alloca double*, align 8
// CHECK9-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK9-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK9-NEXT:    [[VAR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
// CHECK9-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK9-NEXT:    store double* [[G]], double** [[G1]], align 8
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK9-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const.main.vec to i8*), i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR]], i64 0, i64 0
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float 1.000000e+00)
// CHECK9-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_BEGIN]], i64 1
// CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float 2.000000e+00)
// CHECK9-NEXT:    store %struct.S* [[TEST]], %struct.S** [[VAR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S*, %struct.S** [[VAR]], align 8
// CHECK9-NEXT:    store %struct.S* [[TMP1]], %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[T_VAR]], align 4
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* @_ZZ4mainE4svar, align 4
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[SVAR_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP5]], i32* [[CONV1]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[SVAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK9-NEXT:    store i64 [[TMP3]], i64* [[TMP10]], align 8
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK9-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
// CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK9-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK9-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [2 x i32]**
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP15]], align 8
// CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK9-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [2 x i32]**
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP17]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK9-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK9-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [2 x %struct.S]**
// CHECK9-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[TMP20]], align 8
// CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK9-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [2 x %struct.S]**
// CHECK9-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[TMP22]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK9-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK9-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to %struct.S**
// CHECK9-NEXT:    store %struct.S* [[TMP7]], %struct.S** [[TMP25]], align 8
// CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK9-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S**
// CHECK9-NEXT:    store %struct.S* [[TMP8]], %struct.S** [[TMP27]], align 8
// CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK9-NEXT:    store i8* null, i8** [[TMP28]], align 8
// CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK9-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
// CHECK9-NEXT:    store i64 [[TMP6]], i64* [[TMP30]], align 8
// CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK9-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
// CHECK9-NEXT:    store i64 [[TMP6]], i64* [[TMP32]], align 8
// CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK9-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2)
// CHECK9-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK9-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK9-NEXT:    br i1 [[TMP37]], 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_l139(i64 [[TMP3]], [2 x i32]* [[VEC]], [2 x %struct.S]* [[S_ARR]], %struct.S* [[TMP4]], i64 [[TMP6]]) #[[ATTR4:[0-9]+]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
// CHECK9-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP38]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE3:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done3:
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK9-NEXT:    [[TMP39:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK9-NEXT:    ret i32 [[TMP39]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK9-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIfEC2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK9-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK9-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK9-NEXT:    call void @_ZN1SIfEC2Ef(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]], float [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139
// CHECK9-SAME: (i64 [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 [[SVAR:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[SVAR]], i64* [[SVAR_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[SVAR_ADDR]] to i32*
// CHECK9-NEXT:    store %struct.S* [[TMP2]], %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S]*, %struct.S*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], [2 x i32]* [[TMP0]], [2 x %struct.S]* [[TMP1]], %struct.S* [[TMP3]], i32* [[CONV1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32* nonnull align 4 dereferenceable(4) [[SVAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S], align 4
// CHECK9-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK9-NEXT:    [[_TMP8:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[SVAR9:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    store i32* [[SVAR]], i32** [[SVAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[SVAR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S* [[TMP3]], %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    store %struct.S* [[TMP5]], %struct.S** [[_TMP1]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK9-NEXT:    store i32 [[TMP6]], i32* [[T_VAR3]], align 4
// CHECK9-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK9-NEXT:    [[TMP8:%.*]] = bitcast [2 x i32]* [[TMP1]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP7]], i8* align 4 [[TMP8]], i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR5]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP9:%.*]] = bitcast [2 x %struct.S]* [[TMP2]] to %struct.S*
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP10]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK9:       omp.arraycpy.body:
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP9]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[TMP11:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK9-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP11]], i8* align 4 [[TMP12]], i64 4, i1 false)
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP10]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK9:       omp.arraycpy.done6:
// CHECK9-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[_TMP1]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[VAR7]] to i8*
// CHECK9-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP14]], i8* align 4 [[TMP15]], i64 4, i1 false)
// CHECK9-NEXT:    store %struct.S* [[VAR7]], %struct.S** [[_TMP8]], align 8
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK9-NEXT:    store i32 [[TMP16]], i32* [[SVAR9]], align 4
// CHECK9-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP18]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP19]], 1
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP20]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP21]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP22]], [[TMP23]]
// CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK9:       omp.inner.for.cond.cleanup:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[TMP27:%.*]] = zext i32 [[TMP26]] to i64
// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP28]], i32* [[CONV]], align 4
// CHECK9-NEXT:    [[TMP29:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP30:%.*]] = load %struct.S*, %struct.S** [[_TMP8]], align 8
// CHECK9-NEXT:    [[TMP31:%.*]] = load i32, i32* [[SVAR9]], align 4
// CHECK9-NEXT:    [[CONV11:%.*]] = bitcast i64* [[SVAR_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP31]], i32* [[CONV11]], align 4
// CHECK9-NEXT:    [[TMP32:%.*]] = load i64, i64* [[SVAR_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 7, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [2 x i32]*, i64, [2 x %struct.S]*, %struct.S*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP25]], i64 [[TMP27]], [2 x i32]* [[VEC4]], i64 [[TMP29]], [2 x %struct.S]* [[S_ARR5]], %struct.S* [[TMP30]], i64 [[TMP32]])
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP36]])
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN12:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR5]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN12]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP37]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN12]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE13:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done13:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 [[T_VAR:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 [[SVAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC5:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR6:%.*]] = alloca [2 x %struct.S], align 4
// CHECK9-NEXT:    [[VAR8:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK9-NEXT:    [[_TMP9:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[SVAR]], i64* [[SVAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[SVAR_ADDR]] to i32*
// CHECK9-NEXT:    store %struct.S* [[TMP2]], %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK9-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP3]] to i32
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK9-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP4]] to i32
// CHECK9-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = bitcast [2 x i32]* [[VEC5]] to i8*
// CHECK9-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[TMP0]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP5]], i8* align 4 [[TMP6]], i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR6]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP7:%.*]] = bitcast [2 x %struct.S]* [[TMP1]] to %struct.S*
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE7:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK9:       omp.arraycpy.body:
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[TMP9:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK9-NEXT:    [[TMP10:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP9]], i8* align 4 [[TMP10]], i64 4, i1 false)
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE7]], label [[OMP_ARRAYCPY_BODY]]
// CHECK9:       omp.arraycpy.done7:
// CHECK9-NEXT:    [[TMP11:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[VAR8]] to i8*
// CHECK9-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP11]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i64 4, i1 false)
// CHECK9-NEXT:    store %struct.S* [[VAR8]], %struct.S** [[_TMP9]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 1
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK9:       omp.inner.for.cond.cleanup:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP23]] to i64
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC5]], i64 0, i64 [[IDXPROM]]
// CHECK9-NEXT:    store i32 [[TMP22]], i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[_TMP9]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[I]], align 4
// CHECK9-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP25]] to i64
// CHECK9-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR6]], i64 0, i64 [[IDXPROM11]]
// CHECK9-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[ARRAYIDX12]] to i8*
// CHECK9-NEXT:    [[TMP27:%.*]] = bitcast %struct.S* [[TMP24]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP26]], i8* align 4 [[TMP27]], i64 4, i1 false)
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK9-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR8]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN14:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR6]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN14]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN14]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE15:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done15:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK9-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIfED2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK9-SAME: () #[[ATTR5:[0-9]+]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK9-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK9-NEXT:    [[VAR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK9-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiET_v.vec to i8*), i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i64 0, i64 0
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 signext 1)
// CHECK9-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYINIT_BEGIN]], i64 1
// CHECK9-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 signext 2)
// CHECK9-NEXT:    store %struct.S.0* [[TEST]], %struct.S.0** [[VAR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 8
// CHECK9-NEXT:    store %struct.S.0* [[TMP1]], %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[T_VAR]], align 4
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK9-NEXT:    store i64 [[TMP3]], i64* [[TMP8]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK9-NEXT:    store i64 [[TMP3]], i64* [[TMP10]], align 8
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK9-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK9-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to [2 x i32]**
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP13]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK9-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [2 x i32]**
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP15]], align 8
// CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK9-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK9-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [2 x %struct.S.0]**
// CHECK9-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[TMP18]], align 8
// CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK9-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [2 x %struct.S.0]**
// CHECK9-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[TMP20]], align 8
// CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK9-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK9-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to %struct.S.0**
// CHECK9-NEXT:    store %struct.S.0* [[TMP5]], %struct.S.0** [[TMP23]], align 8
// CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK9-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to %struct.S.0**
// CHECK9-NEXT:    store %struct.S.0* [[TMP6]], %struct.S.0** [[TMP25]], align 8
// CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK9-NEXT:    store i8* null, i8** [[TMP26]], align 8
// CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2)
// CHECK9-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// 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_l48(i64 [[TMP3]], [2 x i32]* [[VEC]], [2 x %struct.S.0]* [[S_ARR]], %struct.S.0* [[TMP4]]) #[[ATTR4]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done2:
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK9-NEXT:    [[TMP32:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK9-NEXT:    ret i32 [[TMP32]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK9-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    store float 0.000000e+00, float* [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK9-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK9-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK9-NEXT:    store float [[TMP0]], float* [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK9-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK9-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK9-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIiEC2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK9-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK9-NEXT:    call void @_ZN1SIiEC2Ei(%struct.S.0* 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_l48
// CHECK9-SAME: (i64 [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S.0* [[TMP2]], %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], [2 x i32]* [[TMP0]], [2 x %struct.S.0]* [[TMP1]], %struct.S.0* [[TMP3]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK9-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK9-NEXT:    [[_TMP8:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S.0* [[TMP3]], %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    store %struct.S.0* [[TMP4]], %struct.S.0** [[_TMP1]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK9-NEXT:    store i32 [[TMP5]], i32* [[T_VAR3]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK9-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[TMP1]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP6]], i8* align 4 [[TMP7]], i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP8:%.*]] = bitcast [2 x %struct.S.0]* [[TMP2]] to %struct.S.0*
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP9]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK9:       omp.arraycpy.body:
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[TMP10:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK9-NEXT:    [[TMP11:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP10]], i8* align 4 [[TMP11]], i64 4, i1 false)
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP9]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK9:       omp.arraycpy.done6:
// CHECK9-NEXT:    [[TMP12:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 8
// CHECK9-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[VAR7]] to i8*
// CHECK9-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[TMP12]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK9-NEXT:    store %struct.S.0* [[VAR7]], %struct.S.0** [[_TMP8]], align 8
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP16]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP17]], 1
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK9-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK9:       omp.inner.for.cond.cleanup:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP22]] to i64
// CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP26]], i32* [[CONV]], align 4
// CHECK9-NEXT:    [[TMP27:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8
// CHECK9-NEXT:    [[TMP28:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP8]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [2 x i32]*, i64, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP23]], i64 [[TMP25]], [2 x i32]* [[VEC4]], i64 [[TMP27]], [2 x %struct.S.0]* [[S_ARR5]], %struct.S.0* [[TMP28]])
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], [[TMP30]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP32]])
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN10]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP33]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done11:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 [[T_VAR:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK9-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK9-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK9-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK9-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK9-NEXT:    [[_TMP8:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK9-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8
// CHECK9-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK9-NEXT:    store %struct.S.0* [[TMP2]], %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK9-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP3]] to i32
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK9-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP4]] to i32
// CHECK9-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK9-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[TMP0]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP5]], i8* align 4 [[TMP6]], i64 8, i1 false)
// CHECK9-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP7:%.*]] = bitcast [2 x %struct.S.0]* [[TMP1]] to %struct.S.0*
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 2
// CHECK9-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK9:       omp.arraycpy.body:
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK9-NEXT:    [[TMP9:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK9-NEXT:    [[TMP10:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP9]], i8* align 4 [[TMP10]], i64 4, i1 false)
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK9-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK9:       omp.arraycpy.done6:
// CHECK9-NEXT:    [[TMP11:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK9-NEXT:    [[TMP12:%.*]] = bitcast %struct.S.0* [[VAR7]] to i8*
// CHECK9-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[TMP11]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i64 4, i1 false)
// CHECK9-NEXT:    store %struct.S.0* [[VAR7]], %struct.S.0** [[_TMP8]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 1
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK9-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK9:       omp.inner.for.cond.cleanup:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP23]] to i64
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC4]], i64 0, i64 [[IDXPROM]]
// CHECK9-NEXT:    store i32 [[TMP22]], i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[TMP24:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP8]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[I]], align 4
// CHECK9-NEXT:    [[IDXPROM10:%.*]] = sext i32 [[TMP25]] to i64
// CHECK9-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i64 0, i64 [[IDXPROM10]]
// CHECK9-NEXT:    [[TMP26:%.*]] = bitcast %struct.S.0* [[ARRAYIDX11]] to i8*
// CHECK9-NEXT:    [[TMP27:%.*]] = bitcast %struct.S.0* [[TMP24]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP26]], i8* align 4 [[TMP27]], i64 4, i1 false)
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK9-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAY_BEGIN13:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN13]], i64 2
// CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK9:       arraydestroy.body:
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK9-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN13]]
// CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE14:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK9:       arraydestroy.done14:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK9-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SIiED2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK9-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    store i32 0, i32* [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK9-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK9-NEXT:    store i32 [[TMP0]], i32* [[F]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK9-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK9-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK9-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@main
// CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK10-NEXT:    [[G1:%.*]] = alloca double*, align 4
// CHECK10-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK10-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK10-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
// CHECK10-NEXT:    [[VAR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK10-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK10-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
// CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK10-NEXT:    store double* [[G]], double** [[G1]], align 4
// CHECK10-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK10-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const.main.vec to i8*), i32 8, i1 false)
// CHECK10-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR]], i32 0, i32 0
// CHECK10-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float 1.000000e+00)
// CHECK10-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_BEGIN]], i32 1
// CHECK10-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float 2.000000e+00)
// CHECK10-NEXT:    store %struct.S* [[TEST]], %struct.S** [[VAR]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load %struct.S*, %struct.S** [[VAR]], align 4
// CHECK10-NEXT:    store %struct.S* [[TMP1]], %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[T_VAR]], align 4
// CHECK10-NEXT:    store i32 [[TMP2]], i32* [[T_VAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* @_ZZ4mainE4svar, align 4
// CHECK10-NEXT:    store i32 [[TMP5]], i32* [[SVAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[SVAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP8:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK10-NEXT:    store i32 [[TMP3]], i32* [[TMP10]], align 4
// CHECK10-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK10-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK10-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK10-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK10-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [2 x i32]**
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP15]], align 4
// CHECK10-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK10-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [2 x i32]**
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP17]], align 4
// CHECK10-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK10-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK10-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [2 x %struct.S]**
// CHECK10-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[TMP20]], align 4
// CHECK10-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK10-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [2 x %struct.S]**
// CHECK10-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[TMP22]], align 4
// CHECK10-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK10-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK10-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK10-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to %struct.S**
// CHECK10-NEXT:    store %struct.S* [[TMP7]], %struct.S** [[TMP25]], align 4
// CHECK10-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK10-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S**
// CHECK10-NEXT:    store %struct.S* [[TMP8]], %struct.S** [[TMP27]], align 4
// CHECK10-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK10-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK10-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK10-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK10-NEXT:    store i32 [[TMP6]], i32* [[TMP30]], align 4
// CHECK10-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK10-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i32*
// CHECK10-NEXT:    store i32 [[TMP6]], i32* [[TMP32]], align 4
// CHECK10-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK10-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK10-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2)
// CHECK10-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK10-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK10-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK10:       omp_offload.failed:
// CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139(i32 [[TMP3]], [2 x i32]* [[VEC]], [2 x %struct.S]* [[S_ARR]], %struct.S* [[TMP4]], i32 [[TMP6]]) #[[ATTR4:[0-9]+]]
// CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK10:       omp_offload.cont:
// CHECK10-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiET_v()
// CHECK10-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i32 2
// CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK10:       arraydestroy.body:
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP38]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK10-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK10:       arraydestroy.done2:
// CHECK10-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK10-NEXT:    [[TMP39:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK10-NEXT:    ret i32 [[TMP39]]
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK10-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    call void @_ZN1SIfEC2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK10-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK10-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK10-NEXT:    call void @_ZN1SIfEC2Ef(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]], float [[TMP0]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139
// CHECK10-SAME: (i32 [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 [[SVAR:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 4
// CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[SVAR]], i32* [[SVAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S* [[TMP2]], %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S]*, %struct.S*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[T_VAR_ADDR]], [2 x i32]* [[TMP0]], [2 x %struct.S]* [[TMP1]], %struct.S* [[TMP3]], i32* [[SVAR_ADDR]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32* nonnull align 4 dereferenceable(4) [[SVAR:%.*]]) #[[ATTR3]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 4
// CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[_TMP1:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK10-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S], align 4
// CHECK10-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK10-NEXT:    [[_TMP8:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[SVAR9:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK10-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    store i32* [[SVAR]], i32** [[SVAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[SVAR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S* [[TMP3]], %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    store %struct.S* [[TMP5]], %struct.S** [[_TMP1]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK10-NEXT:    store i32 [[TMP6]], i32* [[T_VAR3]], align 4
// CHECK10-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK10-NEXT:    [[TMP8:%.*]] = bitcast [2 x i32]* [[TMP1]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP7]], i8* align 4 [[TMP8]], i32 8, i1 false)
// CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR5]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP9:%.*]] = bitcast [2 x %struct.S]* [[TMP2]] to %struct.S*
// CHECK10-NEXT:    [[TMP10:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i32 2
// CHECK10-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP10]]
// CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK10:       omp.arraycpy.body:
// CHECK10-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP9]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK10-NEXT:    [[TMP11:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK10-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP11]], i8* align 4 [[TMP12]], i32 4, i1 false)
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK10-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP10]]
// CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK10:       omp.arraycpy.done6:
// CHECK10-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[_TMP1]], align 4
// CHECK10-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[VAR7]] to i8*
// CHECK10-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP14]], i8* align 4 [[TMP15]], i32 4, i1 false)
// CHECK10-NEXT:    store %struct.S* [[VAR7]], %struct.S** [[_TMP8]], align 4
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK10-NEXT:    store i32 [[TMP16]], i32* [[SVAR9]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP18]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP19]], 1
// CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    br label [[COND_END]]
// CHECK10:       cond.end:
// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP20]], [[COND_FALSE]] ]
// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP21]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP22]], [[TMP23]]
// CHECK10-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK10:       omp.inner.for.cond.cleanup:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP26:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK10-NEXT:    store i32 [[TMP26]], i32* [[T_VAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP28:%.*]] = load %struct.S*, %struct.S** [[_TMP8]], align 4
// CHECK10-NEXT:    [[TMP29:%.*]] = load i32, i32* [[SVAR9]], align 4
// CHECK10-NEXT:    store i32 [[TMP29]], i32* [[SVAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[SVAR_CASTED]], align 4
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 7, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [2 x i32]*, i32, [2 x %struct.S]*, %struct.S*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP24]], i32 [[TMP25]], [2 x i32]* [[VEC4]], i32 [[TMP27]], [2 x %struct.S]* [[S_ARR5]], %struct.S* [[TMP28]], i32 [[TMP30]])
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK10:       omp.loop.exit:
// CHECK10-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP34]])
// CHECK10-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAY_BEGIN11:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR5]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN11]], i32 2
// CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK10:       arraydestroy.body:
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP35]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK10-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN11]]
// CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE12:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK10:       arraydestroy.done12:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 [[T_VAR:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 [[SVAR:%.*]]) #[[ATTR3]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 4
// CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
// CHECK10-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S], align 4
// CHECK10-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK10-NEXT:    [[_TMP6:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK10-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK10-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[SVAR]], i32* [[SVAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S* [[TMP2]], %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = bitcast [2 x i32]* [[VEC2]] to i8*
// CHECK10-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[TMP0]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP5]], i8* align 4 [[TMP6]], i32 8, i1 false)
// CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR3]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP7:%.*]] = bitcast [2 x %struct.S]* [[TMP1]] to %struct.S*
// CHECK10-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i32 2
// CHECK10-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK10:       omp.arraycpy.body:
// CHECK10-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK10-NEXT:    [[TMP9:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK10-NEXT:    [[TMP10:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP9]], i8* align 4 [[TMP10]], i32 4, i1 false)
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK10-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
// CHECK10:       omp.arraycpy.done4:
// CHECK10-NEXT:    [[TMP11:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[VAR5]] to i8*
// CHECK10-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP11]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i32 4, i1 false)
// CHECK10-NEXT:    store %struct.S* [[VAR5]], %struct.S** [[_TMP6]], align 4
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 1
// CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    br label [[COND_END]]
// CHECK10:       cond.end:
// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK10-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK10:       omp.inner.for.cond.cleanup:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC2]], i32 0, i32 [[TMP23]]
// CHECK10-NEXT:    store i32 [[TMP22]], i32* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[_TMP6]], align 4
// CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[I]], align 4
// CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR3]], i32 0, i32 [[TMP25]]
// CHECK10-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[ARRAYIDX8]] to i8*
// CHECK10-NEXT:    [[TMP27:%.*]] = bitcast %struct.S* [[TMP24]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP26]], i8* align 4 [[TMP27]], i32 4, i1 false)
// CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK10:       omp.body.continue:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK10-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK10:       omp.loop.exit:
// CHECK10-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK10-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR3]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN10]], i32 2
// CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK10:       arraydestroy.body:
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK10-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK10:       arraydestroy.done11:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK10-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    call void @_ZN1SIfED2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK10-SAME: () #[[ATTR5:[0-9]+]] comdat {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK10-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK10-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK10-NEXT:    [[VAR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK10-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK10-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
// CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK10-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiET_v.vec to i8*), i32 8, i1 false)
// CHECK10-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i32 0, i32 0
// CHECK10-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 1)
// CHECK10-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYINIT_BEGIN]], i32 1
// CHECK10-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 2)
// CHECK10-NEXT:    store %struct.S.0* [[TEST]], %struct.S.0** [[VAR]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 4
// CHECK10-NEXT:    store %struct.S.0* [[TMP1]], %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[T_VAR]], align 4
// CHECK10-NEXT:    store i32 [[TMP2]], i32* [[T_VAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP6:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK10-NEXT:    store i32 [[TMP3]], i32* [[TMP8]], align 4
// CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK10-NEXT:    store i32 [[TMP3]], i32* [[TMP10]], align 4
// CHECK10-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK10-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK10-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK10-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to [2 x i32]**
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP13]], align 4
// CHECK10-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK10-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [2 x i32]**
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP15]], align 4
// CHECK10-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK10-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK10-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [2 x %struct.S.0]**
// CHECK10-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[TMP18]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK10-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [2 x %struct.S.0]**
// CHECK10-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[TMP20]], align 4
// CHECK10-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK10-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK10-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK10-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to %struct.S.0**
// CHECK10-NEXT:    store %struct.S.0* [[TMP5]], %struct.S.0** [[TMP23]], align 4
// CHECK10-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK10-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to %struct.S.0**
// CHECK10-NEXT:    store %struct.S.0* [[TMP6]], %struct.S.0** [[TMP25]], align 4
// CHECK10-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK10-NEXT:    store i8* null, i8** [[TMP26]], align 4
// CHECK10-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2)
// CHECK10-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK10-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK10-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK10:       omp_offload.failed:
// CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48(i32 [[TMP3]], [2 x i32]* [[VEC]], [2 x %struct.S.0]* [[S_ARR]], %struct.S.0* [[TMP4]]) #[[ATTR4]]
// CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK10:       omp_offload.cont:
// CHECK10-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i32 2
// CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK10:       arraydestroy.body:
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK10-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK10:       arraydestroy.done2:
// CHECK10-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK10-NEXT:    [[TMP32:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK10-NEXT:    ret i32 [[TMP32]]
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK10-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    store float 0.000000e+00, float* [[F]], align 4
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK10-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK10-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK10-NEXT:    store float [[TMP0]], float* [[F]], align 4
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK10-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK10-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK10-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    call void @_ZN1SIiEC2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK10-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK10-NEXT:    call void @_ZN1SIiEC2Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]], i32 [[TMP0]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48
// CHECK10-SAME: (i32 [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 4
// CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S.0* [[TMP2]], %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[T_VAR_ADDR]], [2 x i32]* [[TMP0]], [2 x %struct.S.0]* [[TMP1]], %struct.S.0* [[TMP3]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 4
// CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK10-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK10-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK10-NEXT:    [[_TMP8:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK10-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S.0* [[TMP3]], %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    store %struct.S.0* [[TMP4]], %struct.S.0** [[_TMP1]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK10-NEXT:    store i32 [[TMP5]], i32* [[T_VAR3]], align 4
// CHECK10-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK10-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[TMP1]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 [[TMP7]], i32 8, i1 false)
// CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP8:%.*]] = bitcast [2 x %struct.S.0]* [[TMP2]] to %struct.S.0*
// CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i32 2
// CHECK10-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP9]]
// CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK10:       omp.arraycpy.body:
// CHECK10-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK10-NEXT:    [[TMP10:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK10-NEXT:    [[TMP11:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 [[TMP11]], i32 4, i1 false)
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK10-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP9]]
// CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK10:       omp.arraycpy.done6:
// CHECK10-NEXT:    [[TMP12:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[VAR7]] to i8*
// CHECK10-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[TMP12]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i32 4, i1 false)
// CHECK10-NEXT:    store %struct.S.0* [[VAR7]], %struct.S.0** [[_TMP8]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP16]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP17]], 1
// CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    br label [[COND_END]]
// CHECK10:       cond.end:
// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK10-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK10:       omp.inner.for.cond.cleanup:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK10-NEXT:    store i32 [[TMP24]], i32* [[T_VAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4
// CHECK10-NEXT:    [[TMP26:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP8]], align 4
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [2 x i32]*, i32, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP22]], i32 [[TMP23]], [2 x i32]* [[VEC4]], i32 [[TMP25]], [2 x %struct.S.0]* [[S_ARR5]], %struct.S.0* [[TMP26]])
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK10:       omp.loop.exit:
// CHECK10-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK10-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN10]], i32 2
// CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK10:       arraydestroy.body:
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK10-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK10:       arraydestroy.done11:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 [[T_VAR:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 4
// CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
// CHECK10-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK10-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK10-NEXT:    [[_TMP6:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK10-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK10-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK10-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 4
// CHECK10-NEXT:    store %struct.S.0* [[TMP2]], %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = bitcast [2 x i32]* [[VEC2]] to i8*
// CHECK10-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[TMP0]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP5]], i8* align 4 [[TMP6]], i32 8, i1 false)
// CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR3]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP7:%.*]] = bitcast [2 x %struct.S.0]* [[TMP1]] to %struct.S.0*
// CHECK10-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i32 2
// CHECK10-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK10:       omp.arraycpy.body:
// CHECK10-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK10-NEXT:    [[TMP9:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK10-NEXT:    [[TMP10:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP9]], i8* align 4 [[TMP10]], i32 4, i1 false)
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK10-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK10-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
// CHECK10:       omp.arraycpy.done4:
// CHECK10-NEXT:    [[TMP11:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK10-NEXT:    [[TMP12:%.*]] = bitcast %struct.S.0* [[VAR5]] to i8*
// CHECK10-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[TMP11]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i32 4, i1 false)
// CHECK10-NEXT:    store %struct.S.0* [[VAR5]], %struct.S.0** [[_TMP6]], align 4
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 1
// CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    br label [[COND_END]]
// CHECK10:       cond.end:
// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK10-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK10:       omp.inner.for.cond.cleanup:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[T_VAR_ADDR]], align 4
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC2]], i32 0, i32 [[TMP23]]
// CHECK10-NEXT:    store i32 [[TMP22]], i32* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[TMP24:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP6]], align 4
// CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[I]], align 4
// CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR3]], i32 0, i32 [[TMP25]]
// CHECK10-NEXT:    [[TMP26:%.*]] = bitcast %struct.S.0* [[ARRAYIDX8]] to i8*
// CHECK10-NEXT:    [[TMP27:%.*]] = bitcast %struct.S.0* [[TMP24]] to i8*
// CHECK10-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP26]], i8* align 4 [[TMP27]], i32 4, i1 false)
// CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK10:       omp.body.continue:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK10-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK10:       omp.loop.exit:
// CHECK10-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK10-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR3]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN10]], i32 2
// CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK10:       arraydestroy.body:
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK10-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK10:       arraydestroy.done11:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK10-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    call void @_ZN1SIiED2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK10-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    store i32 0, i32* [[F]], align 4
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK10-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP0]], i32* [[F]], align 4
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK10-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK10-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK10-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK10-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK11-NEXT:    [[G1:%.*]] = alloca double*, 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:    [[TMP:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK11-NEXT:    store double* [[G]], double** [[G1]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK11-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const.main.vec to i8*), i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float 1.000000e+00)
// CHECK11-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_BEGIN]], i32 1
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float 2.000000e+00)
// CHECK11-NEXT:    store %struct.S* [[TEST]], %struct.S** [[VAR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S*, %struct.S** [[VAR]], align 4
// CHECK11-NEXT:    store %struct.S* [[TMP1]], %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[T_VAR]], align 4
// CHECK11-NEXT:    store i32 [[TMP2]], i32* [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* @_ZZ4mainE4svar, align 4
// CHECK11-NEXT:    store i32 [[TMP5]], i32* [[SVAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[SVAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[TMP10]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK11-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK11-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [2 x i32]**
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP15]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK11-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [2 x i32]**
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP17]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK11-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [2 x %struct.S]**
// CHECK11-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[TMP20]], align 4
// CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [2 x %struct.S]**
// CHECK11-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[TMP22]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK11-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK11-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to %struct.S**
// CHECK11-NEXT:    store %struct.S* [[TMP7]], %struct.S** [[TMP25]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK11-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S**
// CHECK11-NEXT:    store %struct.S* [[TMP8]], %struct.S** [[TMP27]], align 4
// CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK11-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK11-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK11-NEXT:    store i32 [[TMP6]], i32* [[TMP30]], align 4
// CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK11-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i32*
// CHECK11-NEXT:    store i32 [[TMP6]], i32* [[TMP32]], align 4
// CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK11-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2)
// CHECK11-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK11-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK11-NEXT:    br i1 [[TMP37]], 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_l139(i32 [[TMP3]], [2 x i32]* [[VEC]], [2 x %struct.S]* [[S_ARR]], %struct.S* [[TMP4]], i32 [[TMP6]]) #[[ATTR4:[0-9]+]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK11:       omp_offload.cont:
// CHECK11-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiET_v()
// CHECK11-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP38]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done2:
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK11-NEXT:    [[TMP39:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP39]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK11-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK11-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK11-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC2Ef(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]], float [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139
// CHECK11-SAME: (i32 [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 [[SVAR:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[SVAR]], i32* [[SVAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S* [[TMP2]], %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S]*, %struct.S*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[T_VAR_ADDR]], [2 x i32]* [[TMP0]], [2 x %struct.S]* [[TMP1]], %struct.S* [[TMP3]], i32* [[SVAR_ADDR]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32* nonnull align 4 dereferenceable(4) [[SVAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S], align 4
// CHECK11-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK11-NEXT:    [[_TMP8:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[SVAR9:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32* [[SVAR]], i32** [[SVAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[SVAR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S* [[TMP3]], %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    store %struct.S* [[TMP5]], %struct.S** [[_TMP1]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK11-NEXT:    store i32 [[TMP6]], i32* [[T_VAR3]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK11-NEXT:    [[TMP8:%.*]] = bitcast [2 x i32]* [[TMP1]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP7]], i8* align 4 [[TMP8]], i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR5]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP9:%.*]] = bitcast [2 x %struct.S]* [[TMP2]] to %struct.S*
// CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP10]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK11:       omp.arraycpy.body:
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP9]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[TMP11:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK11-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP11]], i8* align 4 [[TMP12]], i32 4, i1 false)
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP10]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK11:       omp.arraycpy.done6:
// CHECK11-NEXT:    [[TMP13:%.*]] = load %struct.S*, %struct.S** [[_TMP1]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[VAR7]] to i8*
// CHECK11-NEXT:    [[TMP15:%.*]] = bitcast %struct.S* [[TMP13]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP14]], i8* align 4 [[TMP15]], i32 4, i1 false)
// CHECK11-NEXT:    store %struct.S* [[VAR7]], %struct.S** [[_TMP8]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK11-NEXT:    store i32 [[TMP16]], i32* [[SVAR9]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP18]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP19]], 1
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP20]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP21]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP22]], [[TMP23]]
// CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK11:       omp.inner.for.cond.cleanup:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK11-NEXT:    store i32 [[TMP26]], i32* [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP27:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP28:%.*]] = load %struct.S*, %struct.S** [[_TMP8]], align 4
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32, i32* [[SVAR9]], align 4
// CHECK11-NEXT:    store i32 [[TMP29]], i32* [[SVAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[SVAR_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 7, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [2 x i32]*, i32, [2 x %struct.S]*, %struct.S*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP24]], i32 [[TMP25]], [2 x i32]* [[VEC4]], i32 [[TMP27]], [2 x %struct.S]* [[S_ARR5]], %struct.S* [[TMP28]], i32 [[TMP30]])
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP34]])
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN11:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR5]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN11]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP35]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN11]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE12:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done12:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 [[T_VAR:%.*]], [2 x %struct.S]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 [[SVAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S]*, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S], align 4
// CHECK11-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK11-NEXT:    [[_TMP6:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store [2 x %struct.S]* [[S_ARR]], [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[SVAR]], i32* [[SVAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S]*, [2 x %struct.S]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S* [[TMP2]], %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = bitcast [2 x i32]* [[VEC2]] to i8*
// CHECK11-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[TMP0]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP5]], i8* align 4 [[TMP6]], i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR3]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP7:%.*]] = bitcast [2 x %struct.S]* [[TMP1]] to %struct.S*
// CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK11:       omp.arraycpy.body:
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[TMP9:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK11-NEXT:    [[TMP10:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP9]], i8* align 4 [[TMP10]], i32 4, i1 false)
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
// CHECK11:       omp.arraycpy.done4:
// CHECK11-NEXT:    [[TMP11:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[VAR5]] to i8*
// CHECK11-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP11]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i32 4, i1 false)
// CHECK11-NEXT:    store %struct.S* [[VAR5]], %struct.S** [[_TMP6]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 1
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK11-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK11:       omp.inner.for.cond.cleanup:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC2]], i32 0, i32 [[TMP23]]
// CHECK11-NEXT:    store i32 [[TMP22]], i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = load %struct.S*, %struct.S** [[_TMP6]], align 4
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[I]], align 4
// CHECK11-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR3]], i32 0, i32 [[TMP25]]
// CHECK11-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[ARRAYIDX8]] to i8*
// CHECK11-NEXT:    [[TMP27:%.*]] = bitcast %struct.S* [[TMP24]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP26]], i8* align 4 [[TMP27]], i32 4, i1 false)
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK11-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[S_ARR3]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN10]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done11:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK11-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfED2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK11-SAME: () #[[ATTR5:[0-9]+]] comdat {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK11-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK11-NEXT:    [[VAR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK11-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiET_v.vec to i8*), i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 1)
// CHECK11-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYINIT_BEGIN]], i32 1
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 2)
// CHECK11-NEXT:    store %struct.S.0* [[TEST]], %struct.S.0** [[VAR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 4
// CHECK11-NEXT:    store %struct.S.0* [[TMP1]], %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[T_VAR]], align 4
// CHECK11-NEXT:    store i32 [[TMP2]], i32* [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[TMP8]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[TMP10]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK11-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK11-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to [2 x i32]**
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP13]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK11-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [2 x i32]**
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[TMP15]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK11-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [2 x %struct.S.0]**
// CHECK11-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[TMP18]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [2 x %struct.S.0]**
// CHECK11-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[TMP20]], align 4
// CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK11-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK11-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to %struct.S.0**
// CHECK11-NEXT:    store %struct.S.0* [[TMP5]], %struct.S.0** [[TMP23]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK11-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to %struct.S.0**
// CHECK11-NEXT:    store %struct.S.0* [[TMP6]], %struct.S.0** [[TMP25]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK11-NEXT:    store i8* null, i8** [[TMP26]], align 4
// CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2)
// CHECK11-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// 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_l48(i32 [[TMP3]], [2 x i32]* [[VEC]], [2 x %struct.S.0]* [[S_ARR]], %struct.S.0* [[TMP4]]) #[[ATTR4]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK11:       omp_offload.cont:
// CHECK11-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done2:
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
// CHECK11-NEXT:    [[TMP32:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP32]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK11-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    store float 0.000000e+00, float* [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK11-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK11-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK11-NEXT:    store float [[TMP0]], float* [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK11-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 4
// CHECK11-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK11-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiEC2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK11-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiEC2Ei(%struct.S.0* 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_l48
// CHECK11-SAME: (i32 [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S.0* [[TMP2]], %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[T_VAR_ADDR]], [2 x i32]* [[TMP0]], [2 x %struct.S.0]* [[TMP1]], %struct.S.0* [[TMP3]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK11-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK11-NEXT:    [[_TMP8:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S.0* [[TMP3]], %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    store %struct.S.0* [[TMP4]], %struct.S.0** [[_TMP1]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], i32* [[T_VAR3]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[VEC4]] to i8*
// CHECK11-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[TMP1]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 [[TMP7]], i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP8:%.*]] = bitcast [2 x %struct.S.0]* [[TMP2]] to %struct.S.0*
// CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP9]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK11:       omp.arraycpy.body:
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[TMP10:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK11-NEXT:    [[TMP11:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 [[TMP11]], i32 4, i1 false)
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP9]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK11:       omp.arraycpy.done6:
// CHECK11-NEXT:    [[TMP12:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[VAR7]] to i8*
// CHECK11-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[TMP12]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i32 4, i1 false)
// CHECK11-NEXT:    store %struct.S.0* [[VAR7]], %struct.S.0** [[_TMP8]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP16]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP17]], 1
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK11-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK11:       omp.inner.for.cond.cleanup:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK11-NEXT:    store i32 [[TMP24]], i32* [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP8]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [2 x i32]*, i32, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP22]], i32 [[TMP23]], [2 x i32]* [[VEC4]], i32 [[TMP25]], [2 x %struct.S.0]* [[S_ARR5]], %struct.S.0* [[TMP26]])
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR5]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN10]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done11:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 [[T_VAR:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 4
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 4
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
// CHECK11-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK11-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK11-NEXT:    [[_TMP6:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK11-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 4
// CHECK11-NEXT:    store %struct.S.0* [[TMP2]], %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = bitcast [2 x i32]* [[VEC2]] to i8*
// CHECK11-NEXT:    [[TMP6:%.*]] = bitcast [2 x i32]* [[TMP0]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP5]], i8* align 4 [[TMP6]], i32 8, i1 false)
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR3]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP7:%.*]] = bitcast [2 x %struct.S.0]* [[TMP1]] to %struct.S.0*
// CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i32 2
// CHECK11-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK11:       omp.arraycpy.body:
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[TMP9:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK11-NEXT:    [[TMP10:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP9]], i8* align 4 [[TMP10]], i32 4, i1 false)
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
// CHECK11:       omp.arraycpy.done4:
// CHECK11-NEXT:    [[TMP11:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = bitcast %struct.S.0* [[VAR5]] to i8*
// CHECK11-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[TMP11]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i32 4, i1 false)
// CHECK11-NEXT:    store %struct.S.0* [[VAR5]], %struct.S.0** [[_TMP6]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 1
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK11-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK11:       omp.inner.for.cond.cleanup:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[T_VAR_ADDR]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC2]], i32 0, i32 [[TMP23]]
// CHECK11-NEXT:    store i32 [[TMP22]], i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP6]], align 4
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[I]], align 4
// CHECK11-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR3]], i32 0, i32 [[TMP25]]
// CHECK11-NEXT:    [[TMP26:%.*]] = bitcast %struct.S.0* [[ARRAYIDX8]] to i8*
// CHECK11-NEXT:    [[TMP27:%.*]] = bitcast %struct.S.0* [[TMP24]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP26]], i8* align 4 [[TMP27]], i32 4, i1 false)
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK11-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR3]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN10]], i32 2
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP31]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done11:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK11-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiED2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK11-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    store i32 0, i32* [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK11-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], i32* [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK11-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 4
// CHECK11-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK11-NEXT:    ret void
//
//