// 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 _
// expected-no-diagnostics
#ifndef HEADER
#define HEADER

// Test host codegen.
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK2
// RUN: %clang_cc1 -DCK1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3

// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK4
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK6

// RUN: %clang_cc1 -DCK1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK7
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK8

// RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
// RUN: %clang_cc1 -DCK1 -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 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK10
// RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
// RUN: %clang_cc1 -DCK1 -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 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK12
#ifdef CK1

template <typename T, int X, long long Y>
struct SS{
  T a[X];
  float b;
  int foo(void) {

    #pragma omp target
    #pragma omp teams distribute parallel for simd
    for(int i = 0; i < X; i++) {
      a[i] = (T)0;
    }
    #pragma omp target
    #pragma omp teams distribute parallel for simd schedule(static)
    for(int i = 0; i < X; i++) {
      a[i] = (T)0;
    }
    #pragma omp target
    #pragma omp teams distribute parallel for simd schedule(static, X/2)
    for(int i = 0; i < X; i++) {
      a[i] = (T)0;
    }

    #pragma omp target
    #pragma omp teams distribute parallel for simd schedule(dynamic)
    for(int i = 0; i < X; i++) {
      a[i] = (T)0;
    }

    #pragma omp target
    #pragma omp teams distribute parallel for simd schedule(dynamic, X/2)
    for(int i = 0; i < X; i++) {
      a[i] = (T)0;
    }


    return a[0];
  }
};

int teams_template_struct(void) {
  SS<int, 123, 456> V;
  return V.foo();

}
#endif // CK1

// Test host codegen.
// RUN: %clang_cc1 -DCK2 -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK13

// RUN: %clang_cc1 -DCK2 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK14

// RUN: %clang_cc1 -DCK2 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK2 -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK15
// RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK16


// RUN: %clang_cc1 -DCK2 -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK17
// RUN: %clang_cc1 -DCK2 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK2 -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK18

// RUN: %clang_cc1 -DCK2 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK19
// RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK20

// RUN: %clang_cc1 -DCK2 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK21
// RUN: %clang_cc1 -DCK2 -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 -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK22
// RUN: %clang_cc1 -DCK2 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK23
// RUN: %clang_cc1 -DCK2 -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 -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK24
#ifdef CK2

template <typename T, int n>
int tmain(T argc) {
  T a[n];
  int m = 10;
#pragma omp target
#pragma omp teams distribute parallel for simd
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
#pragma omp target
#pragma omp teams distribute parallel for simd schedule(static)
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
#pragma omp target
#pragma omp teams distribute parallel for simd schedule(static, m)
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
#pragma omp target
#pragma omp teams distribute parallel for simd schedule(dynamic)
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
#pragma omp target
#pragma omp teams distribute parallel for simd schedule(dynamic, m)
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
  return 0;
}

int main (int argc, char **argv) {
  int n = 100;
  int a[n];
  int m = 10;
#pragma omp target
#pragma omp teams distribute parallel for simd
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
#pragma omp target
#pragma omp teams distribute parallel for simd dist_schedule(static)
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
#pragma omp target
#pragma omp teams distribute parallel for simd dist_schedule(static, m)
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
#pragma omp target
#pragma omp teams distribute parallel for simd schedule(dynamic)
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
#pragma omp target
#pragma omp teams distribute parallel for simd schedule(dynamic, m)
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
  return tmain<int, 10>(argc);
}


#endif // CK2
#endif // #ifndef HEADER
// CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK1-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK1-NEXT:    ret i32 [[CALL]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK1-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK1-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
// CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK1-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK1-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
// CHECK1-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK1-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK1-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK1-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK1:       omp_offload.failed7:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK1:       omp_offload.cont8:
// CHECK1-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK1-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK1-NEXT:    [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK1-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK1-NEXT:    br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
// CHECK1:       omp_offload.failed14:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT15]]
// CHECK1:       omp_offload.cont15:
// CHECK1-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
// CHECK1-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK1-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK1-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK1-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK1:       omp_offload.failed21:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK1:       omp_offload.cont22:
// CHECK1-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK1-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 8
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP40]], align 8
// CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK1-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK1-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK1-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK1:       omp_offload.failed28:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK1:       omp_offload.cont29:
// CHECK1-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    ret i32 [[TMP45]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !8
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// 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_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !17
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !23
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK1:       omp.dispatch.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK1:       omp.dispatch.body:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// 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, !llvm.access.group !26
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK1:       omp.dispatch.inc:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK1:       omp.dispatch.end:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK1-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !29
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK1-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK1:       omp.dispatch.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK1:       omp.dispatch.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK1:       omp.dispatch.inc:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK1:       omp.dispatch.end:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !35
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK1-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK1:       omp.dispatch.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK1:       omp.dispatch.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK1:       omp.dispatch.inc:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK1:       omp.dispatch.end:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK2-NEXT:    ret i32 [[CALL]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK2-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
// CHECK2-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK2-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK2-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK2-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK2-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK2-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK2:       omp_offload.failed7:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK2:       omp_offload.cont8:
// CHECK2-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK2-NEXT:    [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK2-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK2-NEXT:    br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
// CHECK2:       omp_offload.failed14:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT15]]
// CHECK2:       omp_offload.cont15:
// CHECK2-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP31]], align 8
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK2-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK2-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK2-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK2:       omp_offload.failed21:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK2:       omp_offload.cont22:
// CHECK2-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
// CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 8
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP40]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK2-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK2-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK2-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK2:       omp_offload.failed28:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK2:       omp_offload.cont29:
// CHECK2-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    ret i32 [[TMP45]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !8
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
// 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 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// 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_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// 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 [[TMP4]])
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !17
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// 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 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
// 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 [[TMP4]])
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !23
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// 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 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK2:       omp.dispatch.cond:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK2:       omp.dispatch.body:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// 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, !llvm.access.group !26
// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK2:       omp.dispatch.inc:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK2:       omp.dispatch.end:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK2-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !29
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// 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 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK2:       omp.dispatch.cond:
// CHECK2-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK2:       omp.dispatch.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK2:       omp.dispatch.inc:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK2:       omp.dispatch.end:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !35
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// 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 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// 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:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK2:       omp.dispatch.cond:
// CHECK2-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK2:       omp.dispatch.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK2:       omp.dispatch.inc:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK2:       omp.dispatch.end:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK2-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK3-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK3-NEXT:    ret i32 [[CALL]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK3-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 8
// CHECK3-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
// CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
// CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
// CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK3-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK3-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK3-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK3-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK3:       omp_offload.failed7:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK3:       omp_offload.cont8:
// CHECK3-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK3-NEXT:    [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK3-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK3-NEXT:    br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
// CHECK3:       omp_offload.failed14:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT15]]
// CHECK3:       omp_offload.cont15:
// CHECK3-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 8
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 8
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i64 0, i64 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP31]], align 8
// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK3-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK3-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK3-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK3:       omp_offload.failed21:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK3:       omp_offload.cont22:
// CHECK3-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 8
// CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP40]], align 8
// CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK3-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK3-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK3-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK3:       omp_offload.failed28:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK3:       omp_offload.cont29:
// CHECK3-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 0
// CHECK3-NEXT:    [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    ret i32 [[TMP45]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !8
// CHECK3-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !8
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !8
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK3-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[CONV1]], 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 i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
// CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
// CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
// 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:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !17
// CHECK3-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !17
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !17
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK3-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[CONV1]], 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 i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !20
// CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !20
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !20
// CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !20
// 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:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK3-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !23
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK3-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[CONV1]], 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 i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK3:       omp.dispatch.cond:
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK3:       omp.dispatch.body:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
// CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// 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:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK3-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK3:       omp.dispatch.inc:
// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK3:       omp.dispatch.end:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK3-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK3-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !29
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK3-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[CONV1]], 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 i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK3-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK3:       omp.dispatch.cond:
// CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK3:       omp.dispatch.body:
// 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, !llvm.access.group !32
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK3-NEXT:    br i1 [[CMP]], 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, !llvm.access.group !32
// 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, !llvm.access.group !32
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !32
// CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK3:       omp.dispatch.inc:
// CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK3:       omp.dispatch.end:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK3-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !35
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK3-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[CONV1]], 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 i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK3-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK3:       omp.dispatch.cond:
// CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK3:       omp.dispatch.body:
// 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, !llvm.access.group !38
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK3-NEXT:    br i1 [[CMP]], 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, !llvm.access.group !38
// 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, !llvm.access.group !38
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !38
// CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK3:       omp.dispatch.inc:
// CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK3:       omp.dispatch.end:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3:       .omp.final.then:
// CHECK3-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK3:       .omp.final.done:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK4-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK4-NEXT:    ret i32 [[CALL]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK4-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 8
// CHECK4-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK4-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
// CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK4-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
// CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
// CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK4-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK4-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
// CHECK4-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
// CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
// CHECK4-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
// CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK4-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK4-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK4-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK4:       omp_offload.failed7:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK4:       omp_offload.cont8:
// CHECK4-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK4-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
// CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK4-NEXT:    store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
// CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK4-NEXT:    [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK4-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK4-NEXT:    br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
// CHECK4:       omp_offload.failed14:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT15]]
// CHECK4:       omp_offload.cont15:
// CHECK4-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
// CHECK4-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 8
// CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
// CHECK4-NEXT:    store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 8
// CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i64 0, i64 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP31]], align 8
// CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK4-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK4-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK4-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK4:       omp_offload.failed21:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK4:       omp_offload.cont22:
// CHECK4-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK4-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
// CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK4-NEXT:    store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 8
// CHECK4-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP40]], align 8
// CHECK4-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK4-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK4-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK4-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK4:       omp_offload.failed28:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK4:       omp_offload.cont29:
// CHECK4-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 0
// CHECK4-NEXT:    [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    ret i32 [[TMP45]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !8
// CHECK4-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK4-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !8
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !8
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
// 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 [[TMP2]])
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK4-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[CONV1]], 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 i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
// CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
// CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK4-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
// 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:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// 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 [[TMP4]])
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK4-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !17
// CHECK4-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK4-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !17
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !17
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// 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 [[TMP2]])
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK4-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[CONV1]], 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 i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !20
// CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !20
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !20
// CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK4-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !20
// 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:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
// 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 [[TMP4]])
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK4-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK4-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK4-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !23
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// 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 [[TMP2]])
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK4-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[CONV1]], 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 i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK4:       omp.dispatch.cond:
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
// CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4:       cond.true:
// CHECK4-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK4-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK4:       omp.dispatch.body:
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK4-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
// CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK4-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// 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:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK4-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK4:       omp.dispatch.inc:
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK4-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK4-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK4:       omp.dispatch.end:
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK4-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK4-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK4-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !29
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// 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 [[TMP2]])
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK4-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[CONV1]], 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 i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK4-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK4:       omp.dispatch.cond:
// CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK4:       omp.dispatch.body:
// 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, !llvm.access.group !32
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK4-NEXT:    br i1 [[CMP]], 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, !llvm.access.group !32
// 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, !llvm.access.group !32
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !32
// CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK4-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK4-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK4:       omp.dispatch.inc:
// CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK4:       omp.dispatch.end:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK4-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK4-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !35
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// 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 [[TMP2]])
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK4-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[CONV1]], 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 i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK4-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK4:       omp.dispatch.cond:
// CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK4:       omp.dispatch.body:
// 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, !llvm.access.group !38
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK4-NEXT:    br i1 [[CMP]], 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, !llvm.access.group !38
// 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, !llvm.access.group !38
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !38
// CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK4-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK4-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK4:       omp.dispatch.inc:
// CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK4:       omp.dispatch.end:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4:       .omp.final.then:
// CHECK4-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK4:       .omp.final.done:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK4-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK5-NEXT:    [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK5-NEXT:    ret i32 [[CALL]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK5-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK5-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
// CHECK5-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK5-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK5-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK5:       omp_offload.failed:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK5:       omp_offload.cont:
// CHECK5-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK5-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK5-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK5-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK5-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK5:       omp_offload.failed7:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK5:       omp_offload.cont8:
// CHECK5-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
// CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
// CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK5-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK5-NEXT:    [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK5-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK5-NEXT:    br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
// CHECK5:       omp_offload.failed14:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT15]]
// CHECK5:       omp_offload.cont15:
// CHECK5-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 4
// CHECK5-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 4
// CHECK5-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK5-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK5-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK5-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK5-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK5-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK5:       omp_offload.failed21:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK5:       omp_offload.cont22:
// CHECK5-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
// CHECK5-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 4
// CHECK5-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP40]], align 4
// CHECK5-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK5-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK5-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK5-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK5-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK5:       omp_offload.failed28:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK5:       omp_offload.cont29:
// CHECK5-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK5-NEXT:    ret i32 [[TMP45]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !9
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK5-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !18
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK5-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !24
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK5:       omp.dispatch.cond:
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK5:       omp.dispatch.body:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK5:       omp.dispatch.inc:
// CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK5-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK5-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK5:       omp.dispatch.end:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK5-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK5-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !30
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK5-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK5:       omp.dispatch.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK5:       omp.dispatch.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK5:       omp.dispatch.inc:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK5:       omp.dispatch.end:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK5-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !36
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK5-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK5:       omp.dispatch.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK5:       omp.dispatch.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK5:       omp.dispatch.inc:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK5:       omp.dispatch.end:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK5-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK5-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK5-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK6-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK6-NEXT:    [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK6-NEXT:    ret i32 [[CALL]]
//
//
// CHECK6-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK6-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK6-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
// CHECK6-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK6-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK6-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK6:       omp_offload.failed:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK6:       omp_offload.cont:
// CHECK6-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
// CHECK6-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK6-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK6-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK6-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK6-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK6-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK6:       omp_offload.failed7:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK6:       omp_offload.cont8:
// CHECK6-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
// CHECK6-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
// CHECK6-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK6-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK6-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK6-NEXT:    [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK6-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK6-NEXT:    br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
// CHECK6:       omp_offload.failed14:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT15]]
// CHECK6:       omp_offload.cont15:
// CHECK6-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 4
// CHECK6-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 4
// CHECK6-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK6-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK6-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK6-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK6-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK6-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK6:       omp_offload.failed21:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK6:       omp_offload.cont22:
// CHECK6-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
// CHECK6-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 4
// CHECK6-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP40]], align 4
// CHECK6-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK6-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK6-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK6-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK6-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK6:       omp_offload.failed28:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK6:       omp_offload.cont29:
// CHECK6-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK6-NEXT:    ret i32 [[TMP45]]
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !9
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK6-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK6-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !18
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK6-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK6-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !24
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK6:       omp.dispatch.cond:
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK6:       omp.dispatch.body:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK6-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK6-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK6:       omp.dispatch.inc:
// CHECK6-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK6-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK6-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK6:       omp.dispatch.end:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK6-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK6-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !30
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK6-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK6:       omp.dispatch.cond:
// CHECK6-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK6-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK6-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK6:       omp.dispatch.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK6-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK6:       omp.dispatch.inc:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK6:       omp.dispatch.end:
// CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK6-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !36
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK6-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK6:       omp.dispatch.cond:
// CHECK6-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK6-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK6-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK6:       omp.dispatch.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK6-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK6:       omp.dispatch.inc:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK6:       omp.dispatch.end:
// CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK6-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK6-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK6-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK7-NEXT:    [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK7-NEXT:    ret i32 [[CALL]]
//
//
// CHECK7-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK7-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK7-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
// CHECK7-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK7-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK7-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK7-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
// CHECK7-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK7-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK7-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK7:       omp_offload.failed:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK7:       omp_offload.cont:
// CHECK7-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
// CHECK7-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
// CHECK7-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK7-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK7-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK7-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK7-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK7-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK7-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK7:       omp_offload.failed7:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK7:       omp_offload.cont8:
// CHECK7-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK7-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
// CHECK7-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
// CHECK7-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
// CHECK7-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK7-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK7-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK7-NEXT:    [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK7-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK7-NEXT:    br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
// CHECK7:       omp_offload.failed14:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT15]]
// CHECK7:       omp_offload.cont15:
// CHECK7-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
// CHECK7-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 4
// CHECK7-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 4
// CHECK7-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i32 0, i32 0
// CHECK7-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK7-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK7-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK7-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK7-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK7-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK7:       omp_offload.failed21:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK7:       omp_offload.cont22:
// CHECK7-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK7-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
// CHECK7-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 4
// CHECK7-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
// CHECK7-NEXT:    store i8* null, i8** [[TMP40]], align 4
// CHECK7-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK7-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK7-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK7-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK7-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK7:       omp_offload.failed28:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK7:       omp_offload.cont29:
// CHECK7-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK7-NEXT:    ret i32 [[TMP45]]
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK7:       cond.true:
// CHECK7-NEXT:    br label [[COND_END:%.*]]
// CHECK7:       cond.false:
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    br label [[COND_END]]
// CHECK7:       cond.end:
// CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !9
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !9
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !9
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK7:       omp.loop.exit:
// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK7:       cond.true:
// CHECK7-NEXT:    br label [[COND_END:%.*]]
// CHECK7:       cond.false:
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    br label [[COND_END]]
// CHECK7:       cond.end:
// CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
// CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK7-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
// CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK7:       omp.body.continue:
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK7-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK7:       omp.loop.exit:
// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK7-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK7:       cond.true:
// CHECK7-NEXT:    br label [[COND_END:%.*]]
// CHECK7:       cond.false:
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    br label [[COND_END]]
// CHECK7:       cond.end:
// CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !18
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !18
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !18
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK7:       omp.loop.exit:
// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK7:       cond.true:
// CHECK7-NEXT:    br label [[COND_END:%.*]]
// CHECK7:       cond.false:
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    br label [[COND_END]]
// CHECK7:       cond.end:
// CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
// CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK7-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
// CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK7:       omp.body.continue:
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK7-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK7:       omp.loop.exit:
// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK7-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK7:       cond.true:
// CHECK7-NEXT:    br label [[COND_END:%.*]]
// CHECK7:       cond.false:
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    br label [[COND_END]]
// CHECK7:       cond.end:
// CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !24
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK7:       omp.loop.exit:
// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK7:       omp.dispatch.cond:
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
// CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK7:       cond.true:
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    br label [[COND_END:%.*]]
// CHECK7:       cond.false:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    br label [[COND_END]]
// CHECK7:       cond.end:
// CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK7:       omp.dispatch.body:
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
// CHECK7-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK7-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !27
// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
// CHECK7-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !27
// CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK7:       omp.body.continue:
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK7-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK7-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK7:       omp.dispatch.inc:
// CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK7-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK7-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK7:       omp.dispatch.end:
// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK7-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK7-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK7:       cond.true:
// CHECK7-NEXT:    br label [[COND_END:%.*]]
// CHECK7:       cond.false:
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    br label [[COND_END]]
// CHECK7:       cond.end:
// CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !30
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK7:       omp.loop.exit:
// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK7-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK7:       omp.dispatch.cond:
// CHECK7-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK7-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK7:       omp.dispatch.body:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !33
// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK7-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
// CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK7:       omp.body.continue:
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK7:       omp.dispatch.inc:
// CHECK7-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK7:       omp.dispatch.end:
// CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK7-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK7:       cond.true:
// CHECK7-NEXT:    br label [[COND_END:%.*]]
// CHECK7:       cond.false:
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    br label [[COND_END]]
// CHECK7:       cond.end:
// CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !36
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK7:       omp.loop.exit:
// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK7-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK7-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK7:       omp.dispatch.cond:
// CHECK7-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK7-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK7:       omp.dispatch.body:
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK7-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK7:       omp.inner.for.cond:
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
// CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK7:       omp.inner.for.body:
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !39
// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK7-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !39
// CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK7:       omp.body.continue:
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK7:       omp.inner.for.end:
// CHECK7-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK7:       omp.dispatch.inc:
// CHECK7-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK7:       omp.dispatch.end:
// CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK7-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK7-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK7:       .omp.final.then:
// CHECK7-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK7:       .omp.final.done:
// CHECK7-NEXT:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK7-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK7-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK8-NEXT:    [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK8-NEXT:    ret i32 [[CALL]]
//
//
// CHECK8-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK8-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[_TMP13:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 4
// CHECK8-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK8-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK8-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
// CHECK8-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK8-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 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 123)
// CHECK8-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK8-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK8-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK8:       omp_offload.failed:
// CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK8:       omp_offload.cont:
// CHECK8-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
// CHECK8-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
// CHECK8-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
// CHECK8-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
// CHECK8-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK8-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK8-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK8-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK8-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK8-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK8-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK8:       omp_offload.failed7:
// CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK8:       omp_offload.cont8:
// CHECK8-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK8-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
// CHECK8-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK8-NEXT:    store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
// CHECK8-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
// CHECK8-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK8-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
// CHECK8-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK8-NEXT:    [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK8-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK8-NEXT:    br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
// CHECK8:       omp_offload.failed14:
// CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT15]]
// CHECK8:       omp_offload.cont15:
// CHECK8-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
// CHECK8-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 4
// CHECK8-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
// CHECK8-NEXT:    store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 4
// CHECK8-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i32 0, i32 0
// CHECK8-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK8-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
// CHECK8-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK8-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK8-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK8-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK8:       omp_offload.failed21:
// CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK8:       omp_offload.cont22:
// CHECK8-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK8-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
// CHECK8-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK8-NEXT:    store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 4
// CHECK8-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
// CHECK8-NEXT:    store i8* null, i8** [[TMP40]], align 4
// CHECK8-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK8-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
// CHECK8-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK8-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK8-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK8:       omp_offload.failed28:
// CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK8:       omp_offload.cont29:
// CHECK8-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK8-NEXT:    ret i32 [[TMP45]]
//
//
// CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK8-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK8-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !9
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !9
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !9
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP2]], 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:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
// CHECK8-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK8-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
// CHECK8-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK8-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
// 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:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK8-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK8-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK8-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK8-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK8-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !18
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !18
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !18
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP2]], 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:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// 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:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
// CHECK8-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK8-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
// CHECK8-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK8-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
// 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:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK8-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK8-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK8-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK8-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK8-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !24
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP2]], 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:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK8-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK8:       omp.dispatch.cond:
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
// CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK8:       cond.true:
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    br label [[COND_END:%.*]]
// CHECK8:       cond.false:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK8-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK8:       omp.dispatch.body:
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
// CHECK8-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK8-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
// CHECK8-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !27
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
// CHECK8-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !27
// 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:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK8-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK8-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK8:       omp.dispatch.inc:
// CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK8-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK8-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK8-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK8-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK8:       omp.dispatch.end:
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK8-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK8-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK8-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK8-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !30
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP2]], 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:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK8-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK8-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK8:       omp.dispatch.cond:
// CHECK8-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK8-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK8-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK8:       omp.dispatch.body:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
// CHECK8-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !33
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK8-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK8-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK8-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK8:       omp.dispatch.inc:
// CHECK8-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK8:       omp.dispatch.end:
// CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK8-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 122, 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:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// 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:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    br label [[COND_END]]
// CHECK8:       cond.end:
// CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK8-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK8-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !36
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK8:       omp.inner.for.inc:
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK8:       omp.loop.exit:
// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK8-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP2]], 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:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK8-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK8-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK8:       omp.dispatch.cond:
// CHECK8-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK8-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK8-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK8:       omp.dispatch.body:
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK8-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK8:       omp.inner.for.cond:
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
// CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK8:       omp.inner.for.body:
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
// CHECK8-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !39
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK8-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !39
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK8-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK8-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK8:       omp.inner.for.end:
// CHECK8-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK8:       omp.dispatch.inc:
// CHECK8-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK8:       omp.dispatch.end:
// CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK8-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK8-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK8:       .omp.final.then:
// CHECK8-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK8:       .omp.final.done:
// CHECK8-NEXT:    ret void
//
//
// CHECK8-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK8-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK8-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK9-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK9-NEXT:    ret i32 [[CALL]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK9-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I7:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV23:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I24:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP37:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB38:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB39:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV40:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I41:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP54:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB55:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB56:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV57:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I58:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
// 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:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK9-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB5]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
// CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
// CHECK9:       omp.inner.for.cond8:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
// CHECK9:       omp.inner.for.body10:
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK9-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK9-NEXT:    store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[IDXPROM14:%.*]] = sext i32 [[TMP10]] to i64
// CHECK9-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i64 0, i64 [[IDXPROM14]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX15]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE16:%.*]]
// CHECK9:       omp.body.continue16:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC17:%.*]]
// CHECK9:       omp.inner.for.inc17:
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK9:       omp.inner.for.end19:
// CHECK9-NEXT:    store i32 123, i32* [[I7]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB21]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB22]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB21]], align 4
// CHECK9-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV23]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND25:%.*]]
// CHECK9:       omp.inner.for.cond25:
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB22]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[CMP26:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK9-NEXT:    br i1 [[CMP26]], label [[OMP_INNER_FOR_BODY27:%.*]], label [[OMP_INNER_FOR_END36:%.*]]
// CHECK9:       omp.inner.for.body27:
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[MUL28:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK9-NEXT:    [[ADD29:%.*]] = add nsw i32 0, [[MUL28]]
// CHECK9-NEXT:    store i32 [[ADD29]], i32* [[I24]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I24]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[IDXPROM31:%.*]] = sext i32 [[TMP16]] to i64
// CHECK9-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 [[IDXPROM31]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX32]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE33:%.*]]
// CHECK9:       omp.body.continue33:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC34:%.*]]
// CHECK9:       omp.inner.for.inc34:
// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[ADD35:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK9-NEXT:    store i32 [[ADD35]], i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND25]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK9:       omp.inner.for.end36:
// CHECK9-NEXT:    store i32 123, i32* [[I24]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB38]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB39]], align 4
// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB38]], align 4
// CHECK9-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV40]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND42:%.*]]
// CHECK9:       omp.inner.for.cond42:
// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB39]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[CMP43:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK9-NEXT:    br i1 [[CMP43]], label [[OMP_INNER_FOR_BODY44:%.*]], label [[OMP_INNER_FOR_END53:%.*]]
// CHECK9:       omp.inner.for.body44:
// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[MUL45:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK9-NEXT:    [[ADD46:%.*]] = add nsw i32 0, [[MUL45]]
// CHECK9-NEXT:    store i32 [[ADD46]], i32* [[I41]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[A47:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I41]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[IDXPROM48:%.*]] = sext i32 [[TMP22]] to i64
// CHECK9-NEXT:    [[ARRAYIDX49:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A47]], i64 0, i64 [[IDXPROM48]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX49]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE50:%.*]]
// CHECK9:       omp.body.continue50:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC51:%.*]]
// CHECK9:       omp.inner.for.inc51:
// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[ADD52:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK9-NEXT:    store i32 [[ADD52]], i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND42]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK9:       omp.inner.for.end53:
// CHECK9-NEXT:    store i32 123, i32* [[I41]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB55]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB56]], align 4
// CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB55]], align 4
// CHECK9-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV57]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND59:%.*]]
// CHECK9:       omp.inner.for.cond59:
// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB56]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[CMP60:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK9-NEXT:    br i1 [[CMP60]], label [[OMP_INNER_FOR_BODY61:%.*]], label [[OMP_INNER_FOR_END70:%.*]]
// CHECK9:       omp.inner.for.body61:
// CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[MUL62:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK9-NEXT:    [[ADD63:%.*]] = add nsw i32 0, [[MUL62]]
// CHECK9-NEXT:    store i32 [[ADD63]], i32* [[I58]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[A64:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[I58]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[IDXPROM65:%.*]] = sext i32 [[TMP28]] to i64
// CHECK9-NEXT:    [[ARRAYIDX66:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A64]], i64 0, i64 [[IDXPROM65]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX66]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE67:%.*]]
// CHECK9:       omp.body.continue67:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC68:%.*]]
// CHECK9:       omp.inner.for.inc68:
// CHECK9-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[ADD69:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK9-NEXT:    store i32 [[ADD69]], i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND59]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK9:       omp.inner.for.end70:
// CHECK9-NEXT:    store i32 123, i32* [[I58]], align 4
// CHECK9-NEXT:    [[A71:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[ARRAYIDX72:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A71]], i64 0, i64 0
// CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[ARRAYIDX72]], align 4
// CHECK9-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK10-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK10-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK10-NEXT:    ret i32 [[CALL]]
//
//
// CHECK10-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK10-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I7:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IV23:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I24:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[_TMP37:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB38:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB39:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IV40:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I41:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[_TMP54:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB55:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB56:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IV57:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I58:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
// CHECK10-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
// CHECK10-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
// CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK10-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
// 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:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK10-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK10-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
// CHECK10-NEXT:    store i32 122, i32* [[DOTOMP_UB5]], align 4
// CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
// CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
// CHECK10:       omp.inner.for.cond8:
// CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !6
// CHECK10-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK10-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
// CHECK10:       omp.inner.for.body10:
// CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK10-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK10-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK10-NEXT:    store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !6
// CHECK10-NEXT:    [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !6
// CHECK10-NEXT:    [[IDXPROM14:%.*]] = sext i32 [[TMP10]] to i64
// CHECK10-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i64 0, i64 [[IDXPROM14]]
// CHECK10-NEXT:    store i32 0, i32* [[ARRAYIDX15]], align 4, !llvm.access.group !6
// CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE16:%.*]]
// CHECK10:       omp.body.continue16:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC17:%.*]]
// CHECK10:       omp.inner.for.inc17:
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK10:       omp.inner.for.end19:
// CHECK10-NEXT:    store i32 123, i32* [[I7]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB21]], align 4
// CHECK10-NEXT:    store i32 122, i32* [[DOTOMP_UB22]], align 4
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB21]], align 4
// CHECK10-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV23]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND25:%.*]]
// CHECK10:       omp.inner.for.cond25:
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB22]], align 4, !llvm.access.group !9
// CHECK10-NEXT:    [[CMP26:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK10-NEXT:    br i1 [[CMP26]], label [[OMP_INNER_FOR_BODY27:%.*]], label [[OMP_INNER_FOR_END36:%.*]]
// CHECK10:       omp.inner.for.body27:
// CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK10-NEXT:    [[MUL28:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK10-NEXT:    [[ADD29:%.*]] = add nsw i32 0, [[MUL28]]
// CHECK10-NEXT:    store i32 [[ADD29]], i32* [[I24]], align 4, !llvm.access.group !9
// CHECK10-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I24]], align 4, !llvm.access.group !9
// CHECK10-NEXT:    [[IDXPROM31:%.*]] = sext i32 [[TMP16]] to i64
// CHECK10-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 [[IDXPROM31]]
// CHECK10-NEXT:    store i32 0, i32* [[ARRAYIDX32]], align 4, !llvm.access.group !9
// CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE33:%.*]]
// CHECK10:       omp.body.continue33:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC34:%.*]]
// CHECK10:       omp.inner.for.inc34:
// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK10-NEXT:    [[ADD35:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK10-NEXT:    store i32 [[ADD35]], i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND25]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK10:       omp.inner.for.end36:
// CHECK10-NEXT:    store i32 123, i32* [[I24]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB38]], align 4
// CHECK10-NEXT:    store i32 122, i32* [[DOTOMP_UB39]], align 4
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB38]], align 4
// CHECK10-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV40]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND42:%.*]]
// CHECK10:       omp.inner.for.cond42:
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB39]], align 4, !llvm.access.group !12
// CHECK10-NEXT:    [[CMP43:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK10-NEXT:    br i1 [[CMP43]], label [[OMP_INNER_FOR_BODY44:%.*]], label [[OMP_INNER_FOR_END53:%.*]]
// CHECK10:       omp.inner.for.body44:
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK10-NEXT:    [[MUL45:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK10-NEXT:    [[ADD46:%.*]] = add nsw i32 0, [[MUL45]]
// CHECK10-NEXT:    store i32 [[ADD46]], i32* [[I41]], align 4, !llvm.access.group !12
// CHECK10-NEXT:    [[A47:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I41]], align 4, !llvm.access.group !12
// CHECK10-NEXT:    [[IDXPROM48:%.*]] = sext i32 [[TMP22]] to i64
// CHECK10-NEXT:    [[ARRAYIDX49:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A47]], i64 0, i64 [[IDXPROM48]]
// CHECK10-NEXT:    store i32 0, i32* [[ARRAYIDX49]], align 4, !llvm.access.group !12
// CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE50:%.*]]
// CHECK10:       omp.body.continue50:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC51:%.*]]
// CHECK10:       omp.inner.for.inc51:
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK10-NEXT:    [[ADD52:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK10-NEXT:    store i32 [[ADD52]], i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND42]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK10:       omp.inner.for.end53:
// CHECK10-NEXT:    store i32 123, i32* [[I41]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB55]], align 4
// CHECK10-NEXT:    store i32 122, i32* [[DOTOMP_UB56]], align 4
// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB55]], align 4
// CHECK10-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV57]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND59:%.*]]
// CHECK10:       omp.inner.for.cond59:
// CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK10-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB56]], align 4, !llvm.access.group !15
// CHECK10-NEXT:    [[CMP60:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK10-NEXT:    br i1 [[CMP60]], label [[OMP_INNER_FOR_BODY61:%.*]], label [[OMP_INNER_FOR_END70:%.*]]
// CHECK10:       omp.inner.for.body61:
// CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK10-NEXT:    [[MUL62:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK10-NEXT:    [[ADD63:%.*]] = add nsw i32 0, [[MUL62]]
// CHECK10-NEXT:    store i32 [[ADD63]], i32* [[I58]], align 4, !llvm.access.group !15
// CHECK10-NEXT:    [[A64:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[I58]], align 4, !llvm.access.group !15
// CHECK10-NEXT:    [[IDXPROM65:%.*]] = sext i32 [[TMP28]] to i64
// CHECK10-NEXT:    [[ARRAYIDX66:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A64]], i64 0, i64 [[IDXPROM65]]
// CHECK10-NEXT:    store i32 0, i32* [[ARRAYIDX66]], align 4, !llvm.access.group !15
// CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE67:%.*]]
// CHECK10:       omp.body.continue67:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC68:%.*]]
// CHECK10:       omp.inner.for.inc68:
// CHECK10-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK10-NEXT:    [[ADD69:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK10-NEXT:    store i32 [[ADD69]], i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND59]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK10:       omp.inner.for.end70:
// CHECK10-NEXT:    store i32 123, i32* [[I58]], align 4
// CHECK10-NEXT:    [[A71:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK10-NEXT:    [[ARRAYIDX72:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A71]], i64 0, i64 0
// CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[ARRAYIDX72]], align 4
// CHECK10-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK11-NEXT:    [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK11-NEXT:    ret i32 [[CALL]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK11-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I7:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP19:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB20:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB21:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV22:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I23:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP35:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB36:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB37:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV38:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I39:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB52:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB53:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV54:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I55:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP4]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
// 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:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK11-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB5]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
// CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
// CHECK11:       omp.inner.for.cond8:
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK11-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
// CHECK11:       omp.inner.for.body10:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK11-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK11-NEXT:    store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i32 0, i32 [[TMP10]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX14]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
// CHECK11:       omp.body.continue15:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
// CHECK11:       omp.inner.for.inc16:
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK11-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK11:       omp.inner.for.end18:
// CHECK11-NEXT:    store i32 123, i32* [[I7]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB20]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB21]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB20]], align 4
// CHECK11-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV22]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND24:%.*]]
// CHECK11:       omp.inner.for.cond24:
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB21]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[CMP25:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK11-NEXT:    br i1 [[CMP25]], label [[OMP_INNER_FOR_BODY26:%.*]], label [[OMP_INNER_FOR_END34:%.*]]
// CHECK11:       omp.inner.for.body26:
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[MUL27:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK11-NEXT:    [[ADD28:%.*]] = add nsw i32 0, [[MUL27]]
// CHECK11-NEXT:    store i32 [[ADD28]], i32* [[I23]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[A29:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I23]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[ARRAYIDX30:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A29]], i32 0, i32 [[TMP16]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX30]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE31:%.*]]
// CHECK11:       omp.body.continue31:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC32:%.*]]
// CHECK11:       omp.inner.for.inc32:
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[ADD33:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK11-NEXT:    store i32 [[ADD33]], i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND24]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK11:       omp.inner.for.end34:
// CHECK11-NEXT:    store i32 123, i32* [[I23]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB36]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB37]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB36]], align 4
// CHECK11-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV38]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND40:%.*]]
// CHECK11:       omp.inner.for.cond40:
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB37]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[CMP41:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK11-NEXT:    br i1 [[CMP41]], label [[OMP_INNER_FOR_BODY42:%.*]], label [[OMP_INNER_FOR_END50:%.*]]
// CHECK11:       omp.inner.for.body42:
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[MUL43:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK11-NEXT:    [[ADD44:%.*]] = add nsw i32 0, [[MUL43]]
// CHECK11-NEXT:    store i32 [[ADD44]], i32* [[I39]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[A45:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I39]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[ARRAYIDX46:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A45]], i32 0, i32 [[TMP22]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX46]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE47:%.*]]
// CHECK11:       omp.body.continue47:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC48:%.*]]
// CHECK11:       omp.inner.for.inc48:
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[ADD49:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK11-NEXT:    store i32 [[ADD49]], i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND40]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK11:       omp.inner.for.end50:
// CHECK11-NEXT:    store i32 123, i32* [[I39]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB52]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB53]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB52]], align 4
// CHECK11-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV54]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND56:%.*]]
// CHECK11:       omp.inner.for.cond56:
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB53]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[CMP57:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK11-NEXT:    br i1 [[CMP57]], label [[OMP_INNER_FOR_BODY58:%.*]], label [[OMP_INNER_FOR_END66:%.*]]
// CHECK11:       omp.inner.for.body58:
// CHECK11-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[MUL59:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK11-NEXT:    [[ADD60:%.*]] = add nsw i32 0, [[MUL59]]
// CHECK11-NEXT:    store i32 [[ADD60]], i32* [[I55]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[A61:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP28:%.*]] = load i32, i32* [[I55]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[ARRAYIDX62:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A61]], i32 0, i32 [[TMP28]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX62]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE63:%.*]]
// CHECK11:       omp.body.continue63:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC64:%.*]]
// CHECK11:       omp.inner.for.inc64:
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[ADD65:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK11-NEXT:    store i32 [[ADD65]], i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND56]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK11:       omp.inner.for.end66:
// CHECK11-NEXT:    store i32 123, i32* [[I55]], align 4
// CHECK11-NEXT:    [[A67:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[ARRAYIDX68:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A67]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[ARRAYIDX68]], align 4
// CHECK11-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK12-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK12-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK12-NEXT:    [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
// CHECK12-NEXT:    ret i32 [[CALL]]
//
//
// CHECK12-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK12-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I7:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[_TMP19:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_LB20:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_UB21:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IV22:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I23:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[_TMP35:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_LB36:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_UB37:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IV38:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I39:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_LB52:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_UB53:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IV54:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I55:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
// CHECK12-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK12-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK12:       omp.inner.for.body:
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
// CHECK12-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP4]]
// CHECK12-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
// CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK12:       omp.body.continue:
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK12:       omp.inner.for.inc:
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK12-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK12:       omp.inner.for.end:
// CHECK12-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
// CHECK12-NEXT:    store i32 122, i32* [[DOTOMP_UB5]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
// CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
// CHECK12:       omp.inner.for.cond8:
// CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !7
// CHECK12-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK12-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
// CHECK12:       omp.inner.for.body10:
// CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK12-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK12-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK12-NEXT:    store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !7
// CHECK12-NEXT:    [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !7
// CHECK12-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i32 0, i32 [[TMP10]]
// CHECK12-NEXT:    store i32 0, i32* [[ARRAYIDX14]], align 4, !llvm.access.group !7
// CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
// CHECK12:       omp.body.continue15:
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
// CHECK12:       omp.inner.for.inc16:
// CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK12-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK12-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK12:       omp.inner.for.end18:
// CHECK12-NEXT:    store i32 123, i32* [[I7]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB20]], align 4
// CHECK12-NEXT:    store i32 122, i32* [[DOTOMP_UB21]], align 4
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB20]], align 4
// CHECK12-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV22]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND24:%.*]]
// CHECK12:       omp.inner.for.cond24:
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB21]], align 4, !llvm.access.group !10
// CHECK12-NEXT:    [[CMP25:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK12-NEXT:    br i1 [[CMP25]], label [[OMP_INNER_FOR_BODY26:%.*]], label [[OMP_INNER_FOR_END34:%.*]]
// CHECK12:       omp.inner.for.body26:
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK12-NEXT:    [[MUL27:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK12-NEXT:    [[ADD28:%.*]] = add nsw i32 0, [[MUL27]]
// CHECK12-NEXT:    store i32 [[ADD28]], i32* [[I23]], align 4, !llvm.access.group !10
// CHECK12-NEXT:    [[A29:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I23]], align 4, !llvm.access.group !10
// CHECK12-NEXT:    [[ARRAYIDX30:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A29]], i32 0, i32 [[TMP16]]
// CHECK12-NEXT:    store i32 0, i32* [[ARRAYIDX30]], align 4, !llvm.access.group !10
// CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE31:%.*]]
// CHECK12:       omp.body.continue31:
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC32:%.*]]
// CHECK12:       omp.inner.for.inc32:
// CHECK12-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK12-NEXT:    [[ADD33:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK12-NEXT:    store i32 [[ADD33]], i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND24]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK12:       omp.inner.for.end34:
// CHECK12-NEXT:    store i32 123, i32* [[I23]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB36]], align 4
// CHECK12-NEXT:    store i32 122, i32* [[DOTOMP_UB37]], align 4
// CHECK12-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB36]], align 4
// CHECK12-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV38]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND40:%.*]]
// CHECK12:       omp.inner.for.cond40:
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK12-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB37]], align 4, !llvm.access.group !13
// CHECK12-NEXT:    [[CMP41:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK12-NEXT:    br i1 [[CMP41]], label [[OMP_INNER_FOR_BODY42:%.*]], label [[OMP_INNER_FOR_END50:%.*]]
// CHECK12:       omp.inner.for.body42:
// CHECK12-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK12-NEXT:    [[MUL43:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK12-NEXT:    [[ADD44:%.*]] = add nsw i32 0, [[MUL43]]
// CHECK12-NEXT:    store i32 [[ADD44]], i32* [[I39]], align 4, !llvm.access.group !13
// CHECK12-NEXT:    [[A45:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK12-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I39]], align 4, !llvm.access.group !13
// CHECK12-NEXT:    [[ARRAYIDX46:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A45]], i32 0, i32 [[TMP22]]
// CHECK12-NEXT:    store i32 0, i32* [[ARRAYIDX46]], align 4, !llvm.access.group !13
// CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE47:%.*]]
// CHECK12:       omp.body.continue47:
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC48:%.*]]
// CHECK12:       omp.inner.for.inc48:
// CHECK12-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK12-NEXT:    [[ADD49:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK12-NEXT:    store i32 [[ADD49]], i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND40]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK12:       omp.inner.for.end50:
// CHECK12-NEXT:    store i32 123, i32* [[I39]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB52]], align 4
// CHECK12-NEXT:    store i32 122, i32* [[DOTOMP_UB53]], align 4
// CHECK12-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB52]], align 4
// CHECK12-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV54]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND56:%.*]]
// CHECK12:       omp.inner.for.cond56:
// CHECK12-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK12-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB53]], align 4, !llvm.access.group !16
// CHECK12-NEXT:    [[CMP57:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK12-NEXT:    br i1 [[CMP57]], label [[OMP_INNER_FOR_BODY58:%.*]], label [[OMP_INNER_FOR_END66:%.*]]
// CHECK12:       omp.inner.for.body58:
// CHECK12-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK12-NEXT:    [[MUL59:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK12-NEXT:    [[ADD60:%.*]] = add nsw i32 0, [[MUL59]]
// CHECK12-NEXT:    store i32 [[ADD60]], i32* [[I55]], align 4, !llvm.access.group !16
// CHECK12-NEXT:    [[A61:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK12-NEXT:    [[TMP28:%.*]] = load i32, i32* [[I55]], align 4, !llvm.access.group !16
// CHECK12-NEXT:    [[ARRAYIDX62:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A61]], i32 0, i32 [[TMP28]]
// CHECK12-NEXT:    store i32 0, i32* [[ARRAYIDX62]], align 4, !llvm.access.group !16
// CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE63:%.*]]
// CHECK12:       omp.body.continue63:
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC64:%.*]]
// CHECK12:       omp.inner.for.inc64:
// CHECK12-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK12-NEXT:    [[ADD65:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK12-NEXT:    store i32 [[ADD65]], i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND56]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK12:       omp.inner.for.end66:
// CHECK12-NEXT:    store i32 123, i32* [[I55]], align 4
// CHECK12-NEXT:    [[A67:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK12-NEXT:    [[ARRAYIDX68:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A67]], i32 0, i32 0
// CHECK12-NEXT:    [[TMP30:%.*]] = load i32, i32* [[ARRAYIDX68]], align 4
// CHECK12-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK13-LABEL: define {{[^@]+}}@main
// CHECK13-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK13-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK13-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
// CHECK13-NEXT:    [[_TMP9:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_CASTED19:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
// CHECK13-NEXT:    [[_TMP25:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_CASTED34:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS37:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_SIZES39:%.*]] = alloca [3 x i64], align 8
// CHECK13-NEXT:    [[_TMP40:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[M_CASTED49:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_CASTED51:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS53:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS54:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS55:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_SIZES56:%.*]] = alloca [4 x i64], align 8
// CHECK13-NEXT:    [[_TMP57:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_58:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_59:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK13-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK13-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK13-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK13-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK13-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK13-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK13-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK13-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
// CHECK13-NEXT:    store i64 [[TMP4]], i64* [[TMP7]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
// CHECK13-NEXT:    store i64 [[TMP4]], i64* [[TMP9]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK13-NEXT:    store i64 4, i64* [[TMP10]], align 8
// CHECK13-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK13-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP13]], align 8
// CHECK13-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP15]], align 8
// CHECK13-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK13-NEXT:    store i64 8, i64* [[TMP16]], align 8
// CHECK13-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP17]], align 8
// CHECK13-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP19]], align 8
// CHECK13-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP21]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK13-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK13-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP27:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK13-NEXT:    [[TMP30:%.*]] = zext i32 [[ADD]] to i64
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
// CHECK13-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK13-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK13:       omp_offload.failed:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK13:       omp_offload.cont:
// CHECK13-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
// CHECK13-NEXT:    store i32 [[TMP33]], i32* [[CONV4]], align 4
// CHECK13-NEXT:    [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
// CHECK13-NEXT:    [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK13-NEXT:    store i64 [[TMP34]], i64* [[TMP37]], align 8
// CHECK13-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK13-NEXT:    store i64 [[TMP34]], i64* [[TMP39]], align 8
// CHECK13-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK13-NEXT:    store i64 4, i64* [[TMP40]], align 8
// CHECK13-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP41]], align 8
// CHECK13-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP43]], align 8
// CHECK13-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP45]], align 8
// CHECK13-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
// CHECK13-NEXT:    store i64 8, i64* [[TMP46]], align 8
// CHECK13-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP47]], align 8
// CHECK13-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP49]], align 8
// CHECK13-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP51]], align 8
// CHECK13-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
// CHECK13-NEXT:    store i64 [[TMP35]], i64* [[TMP52]], align 8
// CHECK13-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP53]], align 8
// CHECK13-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP57:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK13-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK13-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
// CHECK13-NEXT:    [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
// CHECK13-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
// CHECK13-NEXT:    store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK13-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK13-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
// CHECK13-NEXT:    [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
// CHECK13-NEXT:    [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159.region_id, i32 3, i8** [[TMP54]], i8** [[TMP55]], i64* [[TMP56]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
// CHECK13-NEXT:    br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK13:       omp_offload.failed16:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK13:       omp_offload.cont17:
// CHECK13-NEXT:    [[TMP63:%.*]] = load i32, i32* [[M]], align 4
// CHECK13-NEXT:    [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP63]], i32* [[CONV18]], align 4
// CHECK13-NEXT:    [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK13-NEXT:    [[TMP65:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
// CHECK13-NEXT:    store i32 [[TMP65]], i32* [[CONV20]], align 4
// CHECK13-NEXT:    [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
// CHECK13-NEXT:    [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
// CHECK13-NEXT:    store i64 [[TMP64]], i64* [[TMP69]], align 8
// CHECK13-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
// CHECK13-NEXT:    store i64 [[TMP64]], i64* [[TMP71]], align 8
// CHECK13-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
// CHECK13-NEXT:    store i64 4, i64* [[TMP72]], align 8
// CHECK13-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP73]], align 8
// CHECK13-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
// CHECK13-NEXT:    store i64 [[TMP66]], i64* [[TMP75]], align 8
// CHECK13-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
// CHECK13-NEXT:    store i64 [[TMP66]], i64* [[TMP77]], align 8
// CHECK13-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
// CHECK13-NEXT:    store i64 4, i64* [[TMP78]], align 8
// CHECK13-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP79]], align 8
// CHECK13-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP81]], align 8
// CHECK13-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP83]], align 8
// CHECK13-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
// CHECK13-NEXT:    store i64 8, i64* [[TMP84]], align 8
// CHECK13-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP85]], align 8
// CHECK13-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
// CHECK13-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP87]], align 8
// CHECK13-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
// CHECK13-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP89]], align 8
// CHECK13-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
// CHECK13-NEXT:    store i64 [[TMP67]], i64* [[TMP90]], align 8
// CHECK13-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
// CHECK13-NEXT:    store i8* null, i8** [[TMP91]], align 8
// CHECK13-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP95:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
// CHECK13-NEXT:    [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
// CHECK13-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
// CHECK13-NEXT:    [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
// CHECK13-NEXT:    [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
// CHECK13-NEXT:    store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK13-NEXT:    [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK13-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
// CHECK13-NEXT:    [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
// CHECK13-NEXT:    [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164.region_id, i32 4, i8** [[TMP92]], i8** [[TMP93]], i64* [[TMP94]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
// CHECK13-NEXT:    br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK13:       omp_offload.failed32:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK13:       omp_offload.cont33:
// CHECK13-NEXT:    [[TMP101:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV35:%.*]] = bitcast i64* [[N_CASTED34]] to i32*
// CHECK13-NEXT:    store i32 [[TMP101]], i32* [[CONV35]], align 4
// CHECK13-NEXT:    [[TMP102:%.*]] = load i64, i64* [[N_CASTED34]], align 8
// CHECK13-NEXT:    [[TMP103:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i64*
// CHECK13-NEXT:    store i64 [[TMP102]], i64* [[TMP105]], align 8
// CHECK13-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
// CHECK13-NEXT:    store i64 [[TMP102]], i64* [[TMP107]], align 8
// CHECK13-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
// CHECK13-NEXT:    store i64 4, i64* [[TMP108]], align 8
// CHECK13-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP109]], align 8
// CHECK13-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP111:%.*]] = bitcast i8** [[TMP110]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP111]], align 8
// CHECK13-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP113:%.*]] = bitcast i8** [[TMP112]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP113]], align 8
// CHECK13-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 1
// CHECK13-NEXT:    store i64 8, i64* [[TMP114]], align 8
// CHECK13-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP115]], align 8
// CHECK13-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP117]], align 8
// CHECK13-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP119]], align 8
// CHECK13-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 2
// CHECK13-NEXT:    store i64 [[TMP103]], i64* [[TMP120]], align 8
// CHECK13-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP121]], align 8
// CHECK13-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP125:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP125]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK13-NEXT:    [[TMP126:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK13-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[TMP126]], 0
// CHECK13-NEXT:    [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
// CHECK13-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
// CHECK13-NEXT:    store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK13-NEXT:    [[TMP127:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK13-NEXT:    [[ADD46:%.*]] = add nsw i32 [[TMP127]], 1
// CHECK13-NEXT:    [[TMP128:%.*]] = zext i32 [[ADD46]] to i64
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP128]])
// CHECK13-NEXT:    [[TMP129:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169.region_id, i32 3, i8** [[TMP122]], i8** [[TMP123]], i64* [[TMP124]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP130:%.*]] = icmp ne i32 [[TMP129]], 0
// CHECK13-NEXT:    br i1 [[TMP130]], label [[OMP_OFFLOAD_FAILED47:%.*]], label [[OMP_OFFLOAD_CONT48:%.*]]
// CHECK13:       omp_offload.failed47:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i64 [[TMP102]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT48]]
// CHECK13:       omp_offload.cont48:
// CHECK13-NEXT:    [[TMP131:%.*]] = load i32, i32* [[M]], align 4
// CHECK13-NEXT:    [[CONV50:%.*]] = bitcast i64* [[M_CASTED49]] to i32*
// CHECK13-NEXT:    store i32 [[TMP131]], i32* [[CONV50]], align 4
// CHECK13-NEXT:    [[TMP132:%.*]] = load i64, i64* [[M_CASTED49]], align 8
// CHECK13-NEXT:    [[TMP133:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV52:%.*]] = bitcast i64* [[N_CASTED51]] to i32*
// CHECK13-NEXT:    store i32 [[TMP133]], i32* [[CONV52]], align 4
// CHECK13-NEXT:    [[TMP134:%.*]] = load i64, i64* [[N_CASTED51]], align 8
// CHECK13-NEXT:    [[TMP135:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i64*
// CHECK13-NEXT:    store i64 [[TMP132]], i64* [[TMP137]], align 8
// CHECK13-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i64*
// CHECK13-NEXT:    store i64 [[TMP132]], i64* [[TMP139]], align 8
// CHECK13-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
// CHECK13-NEXT:    store i64 4, i64* [[TMP140]], align 8
// CHECK13-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP141]], align 8
// CHECK13-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i64*
// CHECK13-NEXT:    store i64 [[TMP134]], i64* [[TMP143]], align 8
// CHECK13-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP145:%.*]] = bitcast i8** [[TMP144]] to i64*
// CHECK13-NEXT:    store i64 [[TMP134]], i64* [[TMP145]], align 8
// CHECK13-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 1
// CHECK13-NEXT:    store i64 4, i64* [[TMP146]], align 8
// CHECK13-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP147]], align 8
// CHECK13-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP149]], align 8
// CHECK13-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP151]], align 8
// CHECK13-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 2
// CHECK13-NEXT:    store i64 8, i64* [[TMP152]], align 8
// CHECK13-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP153]], align 8
// CHECK13-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 3
// CHECK13-NEXT:    [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP155]], align 8
// CHECK13-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 3
// CHECK13-NEXT:    [[TMP157:%.*]] = bitcast i8** [[TMP156]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP157]], align 8
// CHECK13-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 3
// CHECK13-NEXT:    store i64 [[TMP135]], i64* [[TMP158]], align 8
// CHECK13-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 3
// CHECK13-NEXT:    store i8* null, i8** [[TMP159]], align 8
// CHECK13-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP163:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP163]], i32* [[DOTCAPTURE_EXPR_58]], align 4
// CHECK13-NEXT:    [[TMP164:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_58]], align 4
// CHECK13-NEXT:    [[SUB60:%.*]] = sub nsw i32 [[TMP164]], 0
// CHECK13-NEXT:    [[DIV61:%.*]] = sdiv i32 [[SUB60]], 1
// CHECK13-NEXT:    [[SUB62:%.*]] = sub nsw i32 [[DIV61]], 1
// CHECK13-NEXT:    store i32 [[SUB62]], i32* [[DOTCAPTURE_EXPR_59]], align 4
// CHECK13-NEXT:    [[TMP165:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_59]], align 4
// CHECK13-NEXT:    [[ADD63:%.*]] = add nsw i32 [[TMP165]], 1
// CHECK13-NEXT:    [[TMP166:%.*]] = zext i32 [[ADD63]] to i64
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP166]])
// CHECK13-NEXT:    [[TMP167:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174.region_id, i32 4, i8** [[TMP160]], i8** [[TMP161]], i64* [[TMP162]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP168:%.*]] = icmp ne i32 [[TMP167]], 0
// CHECK13-NEXT:    br i1 [[TMP168]], label [[OMP_OFFLOAD_FAILED64:%.*]], label [[OMP_OFFLOAD_CONT65:%.*]]
// CHECK13:       omp_offload.failed64:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i64 [[TMP132]], i64 [[TMP134]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT65]]
// CHECK13:       omp_offload.cont65:
// CHECK13-NEXT:    [[TMP169:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK13-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP169]])
// CHECK13-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK13-NEXT:    [[TMP170:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK13-NEXT:    call void @llvm.stackrestore(i8* [[TMP170]])
// CHECK13-NEXT:    [[TMP171:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK13-NEXT:    ret i32 [[TMP171]]
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK13-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK13-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !13
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK13-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK13-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK13-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK13-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK13-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK13-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !22
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK13-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK13-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK13-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK13-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK13-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK13-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK13-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK13-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*, i64, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !28
// CHECK13-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP21]], i32* [[CONV7]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !28
// CHECK13-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, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]]), !llvm.access.group !28
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
// CHECK13-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK13:       cond.true12:
// CHECK13-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    br label [[COND_END14:%.*]]
// CHECK13:       cond.false13:
// CHECK13-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    br label [[COND_END14]]
// CHECK13:       cond.end14:
// CHECK13-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
// CHECK13-NEXT:    store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
// CHECK13-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK13-NEXT:    br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK13-NEXT:    [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
// CHECK13-NEXT:    [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD18]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK13-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK13-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK13-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK13-NEXT:    [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
// CHECK13-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
// CHECK13-NEXT:    [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
// CHECK13-NEXT:    store i32 [[ADD13]], i32* [[I6]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK13-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..8
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK13-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !34
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK13-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK13-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK13-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK13-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK13-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK13-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK13-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK13-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK13-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK13-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK13-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK13-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*, i64, i32*, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !40
// CHECK13-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP20]], i32* [[CONV7]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !40
// CHECK13-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, i32*, i64, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]]), !llvm.access.group !40
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
// CHECK13-NEXT:    br i1 [[TMP27]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK13-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK13-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK13-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK13-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK13-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK13-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK13-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK13-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK13-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
// CHECK13-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
// CHECK13-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK13-NEXT:    store i32 [[ADD12]], i32* [[I6]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK13-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK13-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[M_CASTED19:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS22:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[_TMP24:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK13-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
// CHECK13-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK13-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK13-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK13-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK13:       omp_offload.failed:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK13:       omp_offload.cont:
// CHECK13-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
// CHECK13-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
// CHECK13-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK13-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK13-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK13:       omp_offload.failed5:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK13:       omp_offload.cont6:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[M]], align 4
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP18]], i32* [[CONV]], align 4
// CHECK13-NEXT:    [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK13-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
// CHECK13-NEXT:    store i64 [[TMP19]], i64* [[TMP21]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK13-NEXT:    store i64 [[TMP19]], i64* [[TMP23]], align 8
// CHECK13-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP24]], align 8
// CHECK13-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
// CHECK13-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
// CHECK13-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK13-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK13-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
// CHECK13-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
// CHECK13:       omp_offload.failed11:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
// CHECK13:       omp_offload.cont12:
// CHECK13-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 8
// CHECK13-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
// CHECK13-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK13-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK13-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK13-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK13:       omp_offload.failed17:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK13:       omp_offload.cont18:
// CHECK13-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK13-NEXT:    [[CONV20:%.*]] = bitcast i64* [[M_CASTED19]] to i32*
// CHECK13-NEXT:    store i32 [[TMP43]], i32* [[CONV20]], align 4
// CHECK13-NEXT:    [[TMP44:%.*]] = load i64, i64* [[M_CASTED19]], align 8
// CHECK13-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK13-NEXT:    store i64 [[TMP44]], i64* [[TMP46]], align 8
// CHECK13-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK13-NEXT:    store i64 [[TMP44]], i64* [[TMP48]], align 8
// CHECK13-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK13-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 8
// CHECK13-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 8
// CHECK13-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK13-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK13-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK13-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK13-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
// CHECK13-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED25:%.*]], label [[OMP_OFFLOAD_CONT26:%.*]]
// CHECK13:       omp_offload.failed25:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i64 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT26]]
// CHECK13:       omp_offload.cont26:
// CHECK13-NEXT:    ret i32 0
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK13-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !46
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP47:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP50:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK13-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !52
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK13-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK13-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !58
// CHECK13-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !58
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !58
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK13-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK13-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK13-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !64
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP65:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK13-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK13-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK13-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP68:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK13-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK13-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !70
// CHECK13-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !70
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !70
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP71:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK13-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK13-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK13-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP74:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK13-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK13-SAME: () #[[ATTR5:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK13-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@main
// CHECK14-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK14-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK14-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
// CHECK14-NEXT:    [[_TMP9:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_CASTED19:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
// CHECK14-NEXT:    [[_TMP25:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[N_CASTED34:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS37:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES39:%.*]] = alloca [3 x i64], align 8
// CHECK14-NEXT:    [[_TMP40:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[M_CASTED49:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_CASTED51:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS53:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS54:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS55:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES56:%.*]] = alloca [4 x i64], align 8
// CHECK14-NEXT:    [[_TMP57:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_58:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_59:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK14-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK14-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK14-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK14-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK14-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK14-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK14-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK14-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
// CHECK14-NEXT:    store i64 [[TMP4]], i64* [[TMP7]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
// CHECK14-NEXT:    store i64 [[TMP4]], i64* [[TMP9]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK14-NEXT:    store i64 4, i64* [[TMP10]], align 8
// CHECK14-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK14-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP13]], align 8
// CHECK14-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP15]], align 8
// CHECK14-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK14-NEXT:    store i64 8, i64* [[TMP16]], align 8
// CHECK14-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP17]], align 8
// CHECK14-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP19]], align 8
// CHECK14-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP21]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK14-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK14-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP27:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK14-NEXT:    [[TMP30:%.*]] = zext i32 [[ADD]] to i64
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
// CHECK14-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK14-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK14:       omp_offload.failed:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK14:       omp_offload.cont:
// CHECK14-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
// CHECK14-NEXT:    store i32 [[TMP33]], i32* [[CONV4]], align 4
// CHECK14-NEXT:    [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
// CHECK14-NEXT:    [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK14-NEXT:    store i64 [[TMP34]], i64* [[TMP37]], align 8
// CHECK14-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK14-NEXT:    store i64 [[TMP34]], i64* [[TMP39]], align 8
// CHECK14-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK14-NEXT:    store i64 4, i64* [[TMP40]], align 8
// CHECK14-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP41]], align 8
// CHECK14-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP43]], align 8
// CHECK14-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP45]], align 8
// CHECK14-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
// CHECK14-NEXT:    store i64 8, i64* [[TMP46]], align 8
// CHECK14-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP47]], align 8
// CHECK14-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP49]], align 8
// CHECK14-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP51]], align 8
// CHECK14-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
// CHECK14-NEXT:    store i64 [[TMP35]], i64* [[TMP52]], align 8
// CHECK14-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP53]], align 8
// CHECK14-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP57:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK14-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK14-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
// CHECK14-NEXT:    [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
// CHECK14-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
// CHECK14-NEXT:    store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK14-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK14-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
// CHECK14-NEXT:    [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
// CHECK14-NEXT:    [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159.region_id, i32 3, i8** [[TMP54]], i8** [[TMP55]], i64* [[TMP56]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
// CHECK14-NEXT:    br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK14:       omp_offload.failed16:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK14:       omp_offload.cont17:
// CHECK14-NEXT:    [[TMP63:%.*]] = load i32, i32* [[M]], align 4
// CHECK14-NEXT:    [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP63]], i32* [[CONV18]], align 4
// CHECK14-NEXT:    [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK14-NEXT:    [[TMP65:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
// CHECK14-NEXT:    store i32 [[TMP65]], i32* [[CONV20]], align 4
// CHECK14-NEXT:    [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
// CHECK14-NEXT:    [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
// CHECK14-NEXT:    store i64 [[TMP64]], i64* [[TMP69]], align 8
// CHECK14-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
// CHECK14-NEXT:    store i64 [[TMP64]], i64* [[TMP71]], align 8
// CHECK14-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
// CHECK14-NEXT:    store i64 4, i64* [[TMP72]], align 8
// CHECK14-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP73]], align 8
// CHECK14-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
// CHECK14-NEXT:    store i64 [[TMP66]], i64* [[TMP75]], align 8
// CHECK14-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
// CHECK14-NEXT:    store i64 [[TMP66]], i64* [[TMP77]], align 8
// CHECK14-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
// CHECK14-NEXT:    store i64 4, i64* [[TMP78]], align 8
// CHECK14-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP79]], align 8
// CHECK14-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP81]], align 8
// CHECK14-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP83]], align 8
// CHECK14-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
// CHECK14-NEXT:    store i64 8, i64* [[TMP84]], align 8
// CHECK14-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP85]], align 8
// CHECK14-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
// CHECK14-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP87]], align 8
// CHECK14-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
// CHECK14-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP89]], align 8
// CHECK14-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
// CHECK14-NEXT:    store i64 [[TMP67]], i64* [[TMP90]], align 8
// CHECK14-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
// CHECK14-NEXT:    store i8* null, i8** [[TMP91]], align 8
// CHECK14-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP95:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
// CHECK14-NEXT:    [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
// CHECK14-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
// CHECK14-NEXT:    [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
// CHECK14-NEXT:    [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
// CHECK14-NEXT:    store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK14-NEXT:    [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK14-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
// CHECK14-NEXT:    [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
// CHECK14-NEXT:    [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164.region_id, i32 4, i8** [[TMP92]], i8** [[TMP93]], i64* [[TMP94]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
// CHECK14-NEXT:    br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK14:       omp_offload.failed32:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK14:       omp_offload.cont33:
// CHECK14-NEXT:    [[TMP101:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV35:%.*]] = bitcast i64* [[N_CASTED34]] to i32*
// CHECK14-NEXT:    store i32 [[TMP101]], i32* [[CONV35]], align 4
// CHECK14-NEXT:    [[TMP102:%.*]] = load i64, i64* [[N_CASTED34]], align 8
// CHECK14-NEXT:    [[TMP103:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i64*
// CHECK14-NEXT:    store i64 [[TMP102]], i64* [[TMP105]], align 8
// CHECK14-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
// CHECK14-NEXT:    store i64 [[TMP102]], i64* [[TMP107]], align 8
// CHECK14-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
// CHECK14-NEXT:    store i64 4, i64* [[TMP108]], align 8
// CHECK14-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP109]], align 8
// CHECK14-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP111:%.*]] = bitcast i8** [[TMP110]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP111]], align 8
// CHECK14-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP113:%.*]] = bitcast i8** [[TMP112]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP113]], align 8
// CHECK14-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 1
// CHECK14-NEXT:    store i64 8, i64* [[TMP114]], align 8
// CHECK14-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP115]], align 8
// CHECK14-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP117]], align 8
// CHECK14-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP119]], align 8
// CHECK14-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 2
// CHECK14-NEXT:    store i64 [[TMP103]], i64* [[TMP120]], align 8
// CHECK14-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP121]], align 8
// CHECK14-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP125:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP125]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK14-NEXT:    [[TMP126:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK14-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[TMP126]], 0
// CHECK14-NEXT:    [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
// CHECK14-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
// CHECK14-NEXT:    store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK14-NEXT:    [[TMP127:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK14-NEXT:    [[ADD46:%.*]] = add nsw i32 [[TMP127]], 1
// CHECK14-NEXT:    [[TMP128:%.*]] = zext i32 [[ADD46]] to i64
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP128]])
// CHECK14-NEXT:    [[TMP129:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169.region_id, i32 3, i8** [[TMP122]], i8** [[TMP123]], i64* [[TMP124]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP130:%.*]] = icmp ne i32 [[TMP129]], 0
// CHECK14-NEXT:    br i1 [[TMP130]], label [[OMP_OFFLOAD_FAILED47:%.*]], label [[OMP_OFFLOAD_CONT48:%.*]]
// CHECK14:       omp_offload.failed47:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i64 [[TMP102]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT48]]
// CHECK14:       omp_offload.cont48:
// CHECK14-NEXT:    [[TMP131:%.*]] = load i32, i32* [[M]], align 4
// CHECK14-NEXT:    [[CONV50:%.*]] = bitcast i64* [[M_CASTED49]] to i32*
// CHECK14-NEXT:    store i32 [[TMP131]], i32* [[CONV50]], align 4
// CHECK14-NEXT:    [[TMP132:%.*]] = load i64, i64* [[M_CASTED49]], align 8
// CHECK14-NEXT:    [[TMP133:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV52:%.*]] = bitcast i64* [[N_CASTED51]] to i32*
// CHECK14-NEXT:    store i32 [[TMP133]], i32* [[CONV52]], align 4
// CHECK14-NEXT:    [[TMP134:%.*]] = load i64, i64* [[N_CASTED51]], align 8
// CHECK14-NEXT:    [[TMP135:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i64*
// CHECK14-NEXT:    store i64 [[TMP132]], i64* [[TMP137]], align 8
// CHECK14-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i64*
// CHECK14-NEXT:    store i64 [[TMP132]], i64* [[TMP139]], align 8
// CHECK14-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
// CHECK14-NEXT:    store i64 4, i64* [[TMP140]], align 8
// CHECK14-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP141]], align 8
// CHECK14-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i64*
// CHECK14-NEXT:    store i64 [[TMP134]], i64* [[TMP143]], align 8
// CHECK14-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP145:%.*]] = bitcast i8** [[TMP144]] to i64*
// CHECK14-NEXT:    store i64 [[TMP134]], i64* [[TMP145]], align 8
// CHECK14-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 1
// CHECK14-NEXT:    store i64 4, i64* [[TMP146]], align 8
// CHECK14-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP147]], align 8
// CHECK14-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP149]], align 8
// CHECK14-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP151]], align 8
// CHECK14-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 2
// CHECK14-NEXT:    store i64 8, i64* [[TMP152]], align 8
// CHECK14-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP153]], align 8
// CHECK14-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 3
// CHECK14-NEXT:    [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP155]], align 8
// CHECK14-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 3
// CHECK14-NEXT:    [[TMP157:%.*]] = bitcast i8** [[TMP156]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP157]], align 8
// CHECK14-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 3
// CHECK14-NEXT:    store i64 [[TMP135]], i64* [[TMP158]], align 8
// CHECK14-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 3
// CHECK14-NEXT:    store i8* null, i8** [[TMP159]], align 8
// CHECK14-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP163:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP163]], i32* [[DOTCAPTURE_EXPR_58]], align 4
// CHECK14-NEXT:    [[TMP164:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_58]], align 4
// CHECK14-NEXT:    [[SUB60:%.*]] = sub nsw i32 [[TMP164]], 0
// CHECK14-NEXT:    [[DIV61:%.*]] = sdiv i32 [[SUB60]], 1
// CHECK14-NEXT:    [[SUB62:%.*]] = sub nsw i32 [[DIV61]], 1
// CHECK14-NEXT:    store i32 [[SUB62]], i32* [[DOTCAPTURE_EXPR_59]], align 4
// CHECK14-NEXT:    [[TMP165:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_59]], align 4
// CHECK14-NEXT:    [[ADD63:%.*]] = add nsw i32 [[TMP165]], 1
// CHECK14-NEXT:    [[TMP166:%.*]] = zext i32 [[ADD63]] to i64
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP166]])
// CHECK14-NEXT:    [[TMP167:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174.region_id, i32 4, i8** [[TMP160]], i8** [[TMP161]], i64* [[TMP162]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP168:%.*]] = icmp ne i32 [[TMP167]], 0
// CHECK14-NEXT:    br i1 [[TMP168]], label [[OMP_OFFLOAD_FAILED64:%.*]], label [[OMP_OFFLOAD_CONT65:%.*]]
// CHECK14:       omp_offload.failed64:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i64 [[TMP132]], i64 [[TMP134]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT65]]
// CHECK14:       omp_offload.cont65:
// CHECK14-NEXT:    [[TMP169:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK14-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP169]])
// CHECK14-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK14-NEXT:    [[TMP170:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK14-NEXT:    call void @llvm.stackrestore(i8* [[TMP170]])
// CHECK14-NEXT:    [[TMP171:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK14-NEXT:    ret i32 [[TMP171]]
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK14-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !13
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK14-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK14-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK14-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK14-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK14-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK14-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK14-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK14-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !22
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK14-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK14-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK14-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK14-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK14-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK14-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK14-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK14-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK14-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK14-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*, i64, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !28
// CHECK14-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP21]], i32* [[CONV7]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !28
// CHECK14-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, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]]), !llvm.access.group !28
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
// CHECK14-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK14:       cond.true12:
// CHECK14-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    br label [[COND_END14:%.*]]
// CHECK14:       cond.false13:
// CHECK14-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    br label [[COND_END14]]
// CHECK14:       cond.end14:
// CHECK14-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
// CHECK14-NEXT:    store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
// CHECK14-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK14-NEXT:    br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK14-NEXT:    [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
// CHECK14-NEXT:    [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD18]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK14-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK14-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK14-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK14-NEXT:    [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
// CHECK14-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
// CHECK14-NEXT:    [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
// CHECK14-NEXT:    store i32 [[ADD13]], i32* [[I6]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK14-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..8
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !34
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK14-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK14-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK14-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK14-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK14-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK14-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK14-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK14-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK14-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK14-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK14-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK14-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*, i64, i32*, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !40
// CHECK14-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP20]], i32* [[CONV7]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !40
// CHECK14-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, i32*, i64, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]]), !llvm.access.group !40
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
// CHECK14-NEXT:    br i1 [[TMP27]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK14-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK14-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK14-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK14-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK14-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK14-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK14-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK14-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
// CHECK14-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
// CHECK14-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK14-NEXT:    store i32 [[ADD12]], i32* [[I6]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK14-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK14-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[M_CASTED19:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS22:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[_TMP24:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK14-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK14-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
// CHECK14-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK14-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK14-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK14-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK14:       omp_offload.failed:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK14:       omp_offload.cont:
// CHECK14-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
// CHECK14-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
// CHECK14-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK14-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK14-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK14-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK14:       omp_offload.failed5:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK14:       omp_offload.cont6:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[M]], align 4
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP18]], i32* [[CONV]], align 4
// CHECK14-NEXT:    [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK14-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
// CHECK14-NEXT:    store i64 [[TMP19]], i64* [[TMP21]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK14-NEXT:    store i64 [[TMP19]], i64* [[TMP23]], align 8
// CHECK14-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP24]], align 8
// CHECK14-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
// CHECK14-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
// CHECK14-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK14-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK14-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
// CHECK14-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
// CHECK14:       omp_offload.failed11:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
// CHECK14:       omp_offload.cont12:
// CHECK14-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 8
// CHECK14-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
// CHECK14-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK14-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK14-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK14-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK14:       omp_offload.failed17:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK14:       omp_offload.cont18:
// CHECK14-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK14-NEXT:    [[CONV20:%.*]] = bitcast i64* [[M_CASTED19]] to i32*
// CHECK14-NEXT:    store i32 [[TMP43]], i32* [[CONV20]], align 4
// CHECK14-NEXT:    [[TMP44:%.*]] = load i64, i64* [[M_CASTED19]], align 8
// CHECK14-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK14-NEXT:    store i64 [[TMP44]], i64* [[TMP46]], align 8
// CHECK14-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK14-NEXT:    store i64 [[TMP44]], i64* [[TMP48]], align 8
// CHECK14-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK14-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 8
// CHECK14-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 8
// CHECK14-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK14-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK14-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK14-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK14-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
// CHECK14-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED25:%.*]], label [[OMP_OFFLOAD_CONT26:%.*]]
// CHECK14:       omp_offload.failed25:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i64 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT26]]
// CHECK14:       omp_offload.cont26:
// CHECK14-NEXT:    ret i32 0
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK14-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !46
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP47:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK14-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK14-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK14-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP50:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK14-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK14-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !52
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK14-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK14-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK14-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK14-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK14-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK14-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !58
// CHECK14-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !58
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !58
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK14-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK14-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK14-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !64
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP65:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK14-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK14-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK14-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK14-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK14-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP68:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK14-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK14-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK14-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !70
// CHECK14-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !70
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !70
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP71:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK14-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK14-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK14-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK14-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK14-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP74:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK14-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK14-SAME: () #[[ATTR5:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK14-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@main
// CHECK15-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK15-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK15-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
// CHECK15-NEXT:    [[_TMP9:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_CASTED19:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
// CHECK15-NEXT:    [[_TMP25:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_CASTED34:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS37:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [3 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES39:%.*]] = alloca [3 x i64], align 8
// CHECK15-NEXT:    [[_TMP40:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M_CASTED49:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_CASTED51:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS53:%.*]] = alloca [4 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS54:%.*]] = alloca [4 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS55:%.*]] = alloca [4 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES56:%.*]] = alloca [4 x i64], align 8
// CHECK15-NEXT:    [[_TMP57:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_58:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_59:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK15-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK15-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK15-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK15-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK15-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK15-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK15-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK15-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK15-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
// CHECK15-NEXT:    store i64 [[TMP4]], i64* [[TMP7]], align 8
// CHECK15-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
// CHECK15-NEXT:    store i64 [[TMP4]], i64* [[TMP9]], align 8
// CHECK15-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK15-NEXT:    store i64 4, i64* [[TMP10]], align 8
// CHECK15-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK15-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP13]], align 8
// CHECK15-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP15]], align 8
// CHECK15-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK15-NEXT:    store i64 8, i64* [[TMP16]], align 8
// CHECK15-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP17]], align 8
// CHECK15-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP19]], align 8
// CHECK15-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP21]], align 8
// CHECK15-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK15-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
// CHECK15-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK15-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP27:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK15-NEXT:    [[TMP30:%.*]] = zext i32 [[ADD]] to i64
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
// CHECK15-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK15-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK15:       omp_offload.failed:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK15:       omp_offload.cont:
// CHECK15-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
// CHECK15-NEXT:    store i32 [[TMP33]], i32* [[CONV4]], align 4
// CHECK15-NEXT:    [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
// CHECK15-NEXT:    [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK15-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK15-NEXT:    store i64 [[TMP34]], i64* [[TMP37]], align 8
// CHECK15-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK15-NEXT:    store i64 [[TMP34]], i64* [[TMP39]], align 8
// CHECK15-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK15-NEXT:    store i64 4, i64* [[TMP40]], align 8
// CHECK15-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP41]], align 8
// CHECK15-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP43]], align 8
// CHECK15-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP45]], align 8
// CHECK15-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
// CHECK15-NEXT:    store i64 8, i64* [[TMP46]], align 8
// CHECK15-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP47]], align 8
// CHECK15-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP49]], align 8
// CHECK15-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP51]], align 8
// CHECK15-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
// CHECK15-NEXT:    store i64 [[TMP35]], i64* [[TMP52]], align 8
// CHECK15-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP53]], align 8
// CHECK15-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP57:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK15-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK15-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
// CHECK15-NEXT:    [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
// CHECK15-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
// CHECK15-NEXT:    store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK15-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK15-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
// CHECK15-NEXT:    [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
// CHECK15-NEXT:    [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159.region_id, i32 3, i8** [[TMP54]], i8** [[TMP55]], i64* [[TMP56]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
// CHECK15-NEXT:    br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK15:       omp_offload.failed16:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK15:       omp_offload.cont17:
// CHECK15-NEXT:    [[TMP63:%.*]] = load i32, i32* [[M]], align 4
// CHECK15-NEXT:    [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP63]], i32* [[CONV18]], align 4
// CHECK15-NEXT:    [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK15-NEXT:    [[TMP65:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
// CHECK15-NEXT:    store i32 [[TMP65]], i32* [[CONV20]], align 4
// CHECK15-NEXT:    [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
// CHECK15-NEXT:    [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK15-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
// CHECK15-NEXT:    store i64 [[TMP64]], i64* [[TMP69]], align 8
// CHECK15-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
// CHECK15-NEXT:    store i64 [[TMP64]], i64* [[TMP71]], align 8
// CHECK15-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
// CHECK15-NEXT:    store i64 4, i64* [[TMP72]], align 8
// CHECK15-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP73]], align 8
// CHECK15-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
// CHECK15-NEXT:    store i64 [[TMP66]], i64* [[TMP75]], align 8
// CHECK15-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
// CHECK15-NEXT:    store i64 [[TMP66]], i64* [[TMP77]], align 8
// CHECK15-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
// CHECK15-NEXT:    store i64 4, i64* [[TMP78]], align 8
// CHECK15-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP79]], align 8
// CHECK15-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP81]], align 8
// CHECK15-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP83]], align 8
// CHECK15-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
// CHECK15-NEXT:    store i64 8, i64* [[TMP84]], align 8
// CHECK15-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP85]], align 8
// CHECK15-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
// CHECK15-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP87]], align 8
// CHECK15-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
// CHECK15-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP89]], align 8
// CHECK15-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
// CHECK15-NEXT:    store i64 [[TMP67]], i64* [[TMP90]], align 8
// CHECK15-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
// CHECK15-NEXT:    store i8* null, i8** [[TMP91]], align 8
// CHECK15-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP95:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
// CHECK15-NEXT:    [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
// CHECK15-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
// CHECK15-NEXT:    [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
// CHECK15-NEXT:    [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
// CHECK15-NEXT:    store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK15-NEXT:    [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK15-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
// CHECK15-NEXT:    [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
// CHECK15-NEXT:    [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164.region_id, i32 4, i8** [[TMP92]], i8** [[TMP93]], i64* [[TMP94]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
// CHECK15-NEXT:    br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK15:       omp_offload.failed32:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK15:       omp_offload.cont33:
// CHECK15-NEXT:    [[TMP101:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[CONV35:%.*]] = bitcast i64* [[N_CASTED34]] to i32*
// CHECK15-NEXT:    store i32 [[TMP101]], i32* [[CONV35]], align 4
// CHECK15-NEXT:    [[TMP102:%.*]] = load i64, i64* [[N_CASTED34]], align 8
// CHECK15-NEXT:    [[TMP103:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK15-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i64*
// CHECK15-NEXT:    store i64 [[TMP102]], i64* [[TMP105]], align 8
// CHECK15-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
// CHECK15-NEXT:    store i64 [[TMP102]], i64* [[TMP107]], align 8
// CHECK15-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
// CHECK15-NEXT:    store i64 4, i64* [[TMP108]], align 8
// CHECK15-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP109]], align 8
// CHECK15-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP111:%.*]] = bitcast i8** [[TMP110]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP111]], align 8
// CHECK15-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP113:%.*]] = bitcast i8** [[TMP112]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP113]], align 8
// CHECK15-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 1
// CHECK15-NEXT:    store i64 8, i64* [[TMP114]], align 8
// CHECK15-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP115]], align 8
// CHECK15-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP117]], align 8
// CHECK15-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP119]], align 8
// CHECK15-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 2
// CHECK15-NEXT:    store i64 [[TMP103]], i64* [[TMP120]], align 8
// CHECK15-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP121]], align 8
// CHECK15-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP125:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP125]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK15-NEXT:    [[TMP126:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK15-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[TMP126]], 0
// CHECK15-NEXT:    [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
// CHECK15-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
// CHECK15-NEXT:    store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK15-NEXT:    [[TMP127:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK15-NEXT:    [[ADD46:%.*]] = add nsw i32 [[TMP127]], 1
// CHECK15-NEXT:    [[TMP128:%.*]] = zext i32 [[ADD46]] to i64
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP128]])
// CHECK15-NEXT:    [[TMP129:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169.region_id, i32 3, i8** [[TMP122]], i8** [[TMP123]], i64* [[TMP124]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP130:%.*]] = icmp ne i32 [[TMP129]], 0
// CHECK15-NEXT:    br i1 [[TMP130]], label [[OMP_OFFLOAD_FAILED47:%.*]], label [[OMP_OFFLOAD_CONT48:%.*]]
// CHECK15:       omp_offload.failed47:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i64 [[TMP102]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT48]]
// CHECK15:       omp_offload.cont48:
// CHECK15-NEXT:    [[TMP131:%.*]] = load i32, i32* [[M]], align 4
// CHECK15-NEXT:    [[CONV50:%.*]] = bitcast i64* [[M_CASTED49]] to i32*
// CHECK15-NEXT:    store i32 [[TMP131]], i32* [[CONV50]], align 4
// CHECK15-NEXT:    [[TMP132:%.*]] = load i64, i64* [[M_CASTED49]], align 8
// CHECK15-NEXT:    [[TMP133:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[CONV52:%.*]] = bitcast i64* [[N_CASTED51]] to i32*
// CHECK15-NEXT:    store i32 [[TMP133]], i32* [[CONV52]], align 4
// CHECK15-NEXT:    [[TMP134:%.*]] = load i64, i64* [[N_CASTED51]], align 8
// CHECK15-NEXT:    [[TMP135:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK15-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i64*
// CHECK15-NEXT:    store i64 [[TMP132]], i64* [[TMP137]], align 8
// CHECK15-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i64*
// CHECK15-NEXT:    store i64 [[TMP132]], i64* [[TMP139]], align 8
// CHECK15-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
// CHECK15-NEXT:    store i64 4, i64* [[TMP140]], align 8
// CHECK15-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP141]], align 8
// CHECK15-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i64*
// CHECK15-NEXT:    store i64 [[TMP134]], i64* [[TMP143]], align 8
// CHECK15-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP145:%.*]] = bitcast i8** [[TMP144]] to i64*
// CHECK15-NEXT:    store i64 [[TMP134]], i64* [[TMP145]], align 8
// CHECK15-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 1
// CHECK15-NEXT:    store i64 4, i64* [[TMP146]], align 8
// CHECK15-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP147]], align 8
// CHECK15-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP149]], align 8
// CHECK15-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
// CHECK15-NEXT:    store i64 [[TMP1]], i64* [[TMP151]], align 8
// CHECK15-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 2
// CHECK15-NEXT:    store i64 8, i64* [[TMP152]], align 8
// CHECK15-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP153]], align 8
// CHECK15-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 3
// CHECK15-NEXT:    [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP155]], align 8
// CHECK15-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 3
// CHECK15-NEXT:    [[TMP157:%.*]] = bitcast i8** [[TMP156]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP157]], align 8
// CHECK15-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 3
// CHECK15-NEXT:    store i64 [[TMP135]], i64* [[TMP158]], align 8
// CHECK15-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 3
// CHECK15-NEXT:    store i8* null, i8** [[TMP159]], align 8
// CHECK15-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP163:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP163]], i32* [[DOTCAPTURE_EXPR_58]], align 4
// CHECK15-NEXT:    [[TMP164:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_58]], align 4
// CHECK15-NEXT:    [[SUB60:%.*]] = sub nsw i32 [[TMP164]], 0
// CHECK15-NEXT:    [[DIV61:%.*]] = sdiv i32 [[SUB60]], 1
// CHECK15-NEXT:    [[SUB62:%.*]] = sub nsw i32 [[DIV61]], 1
// CHECK15-NEXT:    store i32 [[SUB62]], i32* [[DOTCAPTURE_EXPR_59]], align 4
// CHECK15-NEXT:    [[TMP165:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_59]], align 4
// CHECK15-NEXT:    [[ADD63:%.*]] = add nsw i32 [[TMP165]], 1
// CHECK15-NEXT:    [[TMP166:%.*]] = zext i32 [[ADD63]] to i64
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP166]])
// CHECK15-NEXT:    [[TMP167:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174.region_id, i32 4, i8** [[TMP160]], i8** [[TMP161]], i64* [[TMP162]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP168:%.*]] = icmp ne i32 [[TMP167]], 0
// CHECK15-NEXT:    br i1 [[TMP168]], label [[OMP_OFFLOAD_FAILED64:%.*]], label [[OMP_OFFLOAD_CONT65:%.*]]
// CHECK15:       omp_offload.failed64:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i64 [[TMP132]], i64 [[TMP134]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT65]]
// CHECK15:       omp_offload.cont65:
// CHECK15-NEXT:    [[TMP169:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK15-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP169]])
// CHECK15-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK15-NEXT:    [[TMP170:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK15-NEXT:    call void @llvm.stackrestore(i8* [[TMP170]])
// CHECK15-NEXT:    [[TMP171:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK15-NEXT:    ret i32 [[TMP171]]
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK15-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK15-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK15-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !13
// CHECK15-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK15-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !13
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !13
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK15-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK15-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK15-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK15-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK15-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !17
// CHECK15-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK15-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !17
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !17
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !17
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK15-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK15-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK15-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK15-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK15-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK15-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK15-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK15-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK15-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK15-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
// CHECK15-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK15-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !22
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK15-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK15-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK15-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK15-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK15-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK15-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK15-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK15-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK15-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK15-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK15-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK15-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK15-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK15-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK15-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK15-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*, i64, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK15-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK15-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !28
// CHECK15-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP21]], i32* [[CONV7]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !28
// CHECK15-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, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]]), !llvm.access.group !28
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK15-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK15-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
// CHECK15-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK15:       cond.true12:
// CHECK15-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    br label [[COND_END14:%.*]]
// CHECK15:       cond.false13:
// CHECK15-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    br label [[COND_END14]]
// CHECK15:       cond.end14:
// CHECK15-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
// CHECK15-NEXT:    store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
// CHECK15-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK15-NEXT:    br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK15-NEXT:    [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
// CHECK15-NEXT:    [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD18]], i32* [[I4]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK15-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK15-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK15-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !31
// CHECK15-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK15-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !31
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !31
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !31
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK15-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK15-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK15-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK15-NEXT:    [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
// CHECK15-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
// CHECK15-NEXT:    [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
// CHECK15-NEXT:    store i32 [[ADD13]], i32* [[I6]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK15-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK15-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..8
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK15-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !34
// CHECK15-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK15-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !34
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK15-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK15-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK15-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK15-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK15-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK15-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK15:       omp.dispatch.cond:
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK15-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK15-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK15:       omp.dispatch.body:
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !37
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !37
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !37
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !37
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK15-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK15-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK15:       omp.dispatch.inc:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK15:       omp.dispatch.end:
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK15-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK15-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK15-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK15-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK15-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK15-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK15-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK15-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK15-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*, i64, i32*, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK15-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK15-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !40
// CHECK15-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK15-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !40
// CHECK15-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP20]], i32* [[CONV7]], align 4, !llvm.access.group !40
// CHECK15-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !40
// CHECK15-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, i32*, i64, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]]), !llvm.access.group !40
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !40
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK15-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
// CHECK15-NEXT:    br i1 [[TMP27]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK15-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK15-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK15-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK15-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK15:       omp.precond.then:
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK15-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK15-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK15-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK15:       omp.dispatch.cond:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK15-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK15-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK15-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK15:       omp.dispatch.body:
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
// CHECK15-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK15-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !43
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !43
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !43
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK15:       omp.dispatch.inc:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK15:       omp.dispatch.end:
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK15-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK15-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
// CHECK15-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
// CHECK15-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK15-NEXT:    store i32 [[ADD12]], i32* [[I6]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK15-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK15-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
// CHECK15-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 8
// CHECK15-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M_CASTED19:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [2 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS22:%.*]] = alloca [2 x i8*], align 8
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [2 x i8*], align 8
// CHECK15-NEXT:    [[_TMP24:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK15-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
// CHECK15-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK15-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK15-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK15-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK15:       omp_offload.failed:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK15:       omp_offload.cont:
// CHECK15-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
// CHECK15-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
// CHECK15-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK15-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK15-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK15-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK15:       omp_offload.failed5:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK15:       omp_offload.cont6:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[M]], align 4
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP18]], i32* [[CONV]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK15-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
// CHECK15-NEXT:    store i64 [[TMP19]], i64* [[TMP21]], align 8
// CHECK15-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK15-NEXT:    store i64 [[TMP19]], i64* [[TMP23]], align 8
// CHECK15-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP24]], align 8
// CHECK15-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
// CHECK15-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
// CHECK15-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK15-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK15-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
// CHECK15-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
// CHECK15:       omp_offload.failed11:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
// CHECK15:       omp_offload.cont12:
// CHECK15-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 8
// CHECK15-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
// CHECK15-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK15-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK15-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK15-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK15:       omp_offload.failed17:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK15:       omp_offload.cont18:
// CHECK15-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK15-NEXT:    [[CONV20:%.*]] = bitcast i64* [[M_CASTED19]] to i32*
// CHECK15-NEXT:    store i32 [[TMP43]], i32* [[CONV20]], align 4
// CHECK15-NEXT:    [[TMP44:%.*]] = load i64, i64* [[M_CASTED19]], align 8
// CHECK15-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK15-NEXT:    store i64 [[TMP44]], i64* [[TMP46]], align 8
// CHECK15-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK15-NEXT:    store i64 [[TMP44]], i64* [[TMP48]], align 8
// CHECK15-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK15-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 8
// CHECK15-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 8
// CHECK15-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK15-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK15-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK15-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK15-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
// CHECK15-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED25:%.*]], label [[OMP_OFFLOAD_CONT26:%.*]]
// CHECK15:       omp_offload.failed25:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i64 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT26]]
// CHECK15:       omp_offload.cont26:
// CHECK15-NEXT:    ret i32 0
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK15-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK15-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK15-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !46
// CHECK15-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK15-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !46
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !46
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP47:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK15-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK15-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK15-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !49
// CHECK15-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK15-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !49
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !49
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !49
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK15-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK15-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP50:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK15-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK15-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK15-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK15-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !52
// CHECK15-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK15-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !52
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !52
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK15-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK15-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK15-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !55
// CHECK15-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK15-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !55
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !55
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !55
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK15-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK15-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK15-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK15-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK15-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK15-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !58
// CHECK15-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK15-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !58
// CHECK15-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !58
// CHECK15-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !58
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !58
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !58
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK15-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK15-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK15-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK15:       omp.dispatch.cond:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK15-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK15:       omp.dispatch.body:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
// CHECK15-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK15-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK15-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK15-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK15:       omp.dispatch.inc:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK15-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK15:       omp.dispatch.end:
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK15-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK15-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK15-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK15-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !64
// CHECK15-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK15-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !64
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !64
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP65:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK15-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK15-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK15-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK15-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK15:       omp.dispatch.cond:
// CHECK15-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK15-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK15-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK15:       omp.dispatch.body:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !67
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !67
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !67
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !67
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK15-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK15-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP68:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK15:       omp.dispatch.inc:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK15:       omp.dispatch.end:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK15-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK15-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    br label [[COND_END:%.*]]
// CHECK15:       cond.false:
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END]]
// CHECK15:       cond.end:
// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK15-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK15-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !70
// CHECK15-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK15-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !70
// CHECK15-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK15-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !70
// CHECK15-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !70
// CHECK15-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !70
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !70
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP71:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK15-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK15-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK15-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK15-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK15-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK15-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK15:       omp.dispatch.cond:
// CHECK15-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK15-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK15-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK15:       omp.dispatch.body:
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK15:       omp.inner.for.cond:
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !73
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK15:       omp.inner.for.body:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !73
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !73
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !73
// CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK15:       omp.body.continue:
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK15-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK15-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP74:![0-9]+]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK15:       omp.dispatch.inc:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK15:       omp.dispatch.end:
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK15-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK15-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK15:       .omp.final.then:
// CHECK15-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK15:       .omp.final.done:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK15-SAME: () #[[ATTR5:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK15-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@main
// CHECK16-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK16-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK16-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
// CHECK16-NEXT:    [[_TMP9:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_CASTED19:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
// CHECK16-NEXT:    [[_TMP25:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[N_CASTED34:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS37:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [3 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_SIZES39:%.*]] = alloca [3 x i64], align 8
// CHECK16-NEXT:    [[_TMP40:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[M_CASTED49:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_CASTED51:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS53:%.*]] = alloca [4 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS54:%.*]] = alloca [4 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS55:%.*]] = alloca [4 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_SIZES56:%.*]] = alloca [4 x i64], align 8
// CHECK16-NEXT:    [[_TMP57:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_58:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_59:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK16-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK16-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK16-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK16-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK16-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK16-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK16-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK16-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK16-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
// CHECK16-NEXT:    store i64 [[TMP4]], i64* [[TMP7]], align 8
// CHECK16-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
// CHECK16-NEXT:    store i64 [[TMP4]], i64* [[TMP9]], align 8
// CHECK16-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK16-NEXT:    store i64 4, i64* [[TMP10]], align 8
// CHECK16-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK16-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP13]], align 8
// CHECK16-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP15]], align 8
// CHECK16-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK16-NEXT:    store i64 8, i64* [[TMP16]], align 8
// CHECK16-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK16-NEXT:    store i8* null, i8** [[TMP17]], align 8
// CHECK16-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP19]], align 8
// CHECK16-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP21]], align 8
// CHECK16-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK16-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
// CHECK16-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK16-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK16-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP27:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK16-NEXT:    [[TMP30:%.*]] = zext i32 [[ADD]] to i64
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
// CHECK16-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK16-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK16:       omp_offload.failed:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK16:       omp_offload.cont:
// CHECK16-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
// CHECK16-NEXT:    store i32 [[TMP33]], i32* [[CONV4]], align 4
// CHECK16-NEXT:    [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
// CHECK16-NEXT:    [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK16-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK16-NEXT:    store i64 [[TMP34]], i64* [[TMP37]], align 8
// CHECK16-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK16-NEXT:    store i64 [[TMP34]], i64* [[TMP39]], align 8
// CHECK16-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK16-NEXT:    store i64 4, i64* [[TMP40]], align 8
// CHECK16-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP41]], align 8
// CHECK16-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP43]], align 8
// CHECK16-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP45]], align 8
// CHECK16-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
// CHECK16-NEXT:    store i64 8, i64* [[TMP46]], align 8
// CHECK16-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
// CHECK16-NEXT:    store i8* null, i8** [[TMP47]], align 8
// CHECK16-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP49]], align 8
// CHECK16-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP51]], align 8
// CHECK16-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
// CHECK16-NEXT:    store i64 [[TMP35]], i64* [[TMP52]], align 8
// CHECK16-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
// CHECK16-NEXT:    store i8* null, i8** [[TMP53]], align 8
// CHECK16-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP57:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK16-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK16-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
// CHECK16-NEXT:    [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
// CHECK16-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
// CHECK16-NEXT:    store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK16-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK16-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
// CHECK16-NEXT:    [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
// CHECK16-NEXT:    [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159.region_id, i32 3, i8** [[TMP54]], i8** [[TMP55]], i64* [[TMP56]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
// CHECK16-NEXT:    br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK16:       omp_offload.failed16:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK16:       omp_offload.cont17:
// CHECK16-NEXT:    [[TMP63:%.*]] = load i32, i32* [[M]], align 4
// CHECK16-NEXT:    [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP63]], i32* [[CONV18]], align 4
// CHECK16-NEXT:    [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK16-NEXT:    [[TMP65:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
// CHECK16-NEXT:    store i32 [[TMP65]], i32* [[CONV20]], align 4
// CHECK16-NEXT:    [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
// CHECK16-NEXT:    [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK16-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
// CHECK16-NEXT:    store i64 [[TMP64]], i64* [[TMP69]], align 8
// CHECK16-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
// CHECK16-NEXT:    store i64 [[TMP64]], i64* [[TMP71]], align 8
// CHECK16-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
// CHECK16-NEXT:    store i64 4, i64* [[TMP72]], align 8
// CHECK16-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP73]], align 8
// CHECK16-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
// CHECK16-NEXT:    store i64 [[TMP66]], i64* [[TMP75]], align 8
// CHECK16-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
// CHECK16-NEXT:    store i64 [[TMP66]], i64* [[TMP77]], align 8
// CHECK16-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
// CHECK16-NEXT:    store i64 4, i64* [[TMP78]], align 8
// CHECK16-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
// CHECK16-NEXT:    store i8* null, i8** [[TMP79]], align 8
// CHECK16-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP81]], align 8
// CHECK16-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP83]], align 8
// CHECK16-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
// CHECK16-NEXT:    store i64 8, i64* [[TMP84]], align 8
// CHECK16-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
// CHECK16-NEXT:    store i8* null, i8** [[TMP85]], align 8
// CHECK16-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
// CHECK16-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP87]], align 8
// CHECK16-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
// CHECK16-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP89]], align 8
// CHECK16-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
// CHECK16-NEXT:    store i64 [[TMP67]], i64* [[TMP90]], align 8
// CHECK16-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
// CHECK16-NEXT:    store i8* null, i8** [[TMP91]], align 8
// CHECK16-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP95:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
// CHECK16-NEXT:    [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
// CHECK16-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
// CHECK16-NEXT:    [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
// CHECK16-NEXT:    [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
// CHECK16-NEXT:    store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK16-NEXT:    [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK16-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
// CHECK16-NEXT:    [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
// CHECK16-NEXT:    [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164.region_id, i32 4, i8** [[TMP92]], i8** [[TMP93]], i64* [[TMP94]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
// CHECK16-NEXT:    br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK16:       omp_offload.failed32:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK16:       omp_offload.cont33:
// CHECK16-NEXT:    [[TMP101:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[CONV35:%.*]] = bitcast i64* [[N_CASTED34]] to i32*
// CHECK16-NEXT:    store i32 [[TMP101]], i32* [[CONV35]], align 4
// CHECK16-NEXT:    [[TMP102:%.*]] = load i64, i64* [[N_CASTED34]], align 8
// CHECK16-NEXT:    [[TMP103:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK16-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i64*
// CHECK16-NEXT:    store i64 [[TMP102]], i64* [[TMP105]], align 8
// CHECK16-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
// CHECK16-NEXT:    store i64 [[TMP102]], i64* [[TMP107]], align 8
// CHECK16-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
// CHECK16-NEXT:    store i64 4, i64* [[TMP108]], align 8
// CHECK16-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP109]], align 8
// CHECK16-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP111:%.*]] = bitcast i8** [[TMP110]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP111]], align 8
// CHECK16-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP113:%.*]] = bitcast i8** [[TMP112]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP113]], align 8
// CHECK16-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 1
// CHECK16-NEXT:    store i64 8, i64* [[TMP114]], align 8
// CHECK16-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
// CHECK16-NEXT:    store i8* null, i8** [[TMP115]], align 8
// CHECK16-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP117]], align 8
// CHECK16-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP119]], align 8
// CHECK16-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 2
// CHECK16-NEXT:    store i64 [[TMP103]], i64* [[TMP120]], align 8
// CHECK16-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 2
// CHECK16-NEXT:    store i8* null, i8** [[TMP121]], align 8
// CHECK16-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP125:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    store i32 [[TMP125]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK16-NEXT:    [[TMP126:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK16-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[TMP126]], 0
// CHECK16-NEXT:    [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
// CHECK16-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
// CHECK16-NEXT:    store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK16-NEXT:    [[TMP127:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK16-NEXT:    [[ADD46:%.*]] = add nsw i32 [[TMP127]], 1
// CHECK16-NEXT:    [[TMP128:%.*]] = zext i32 [[ADD46]] to i64
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP128]])
// CHECK16-NEXT:    [[TMP129:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169.region_id, i32 3, i8** [[TMP122]], i8** [[TMP123]], i64* [[TMP124]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP130:%.*]] = icmp ne i32 [[TMP129]], 0
// CHECK16-NEXT:    br i1 [[TMP130]], label [[OMP_OFFLOAD_FAILED47:%.*]], label [[OMP_OFFLOAD_CONT48:%.*]]
// CHECK16:       omp_offload.failed47:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i64 [[TMP102]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT48]]
// CHECK16:       omp_offload.cont48:
// CHECK16-NEXT:    [[TMP131:%.*]] = load i32, i32* [[M]], align 4
// CHECK16-NEXT:    [[CONV50:%.*]] = bitcast i64* [[M_CASTED49]] to i32*
// CHECK16-NEXT:    store i32 [[TMP131]], i32* [[CONV50]], align 4
// CHECK16-NEXT:    [[TMP132:%.*]] = load i64, i64* [[M_CASTED49]], align 8
// CHECK16-NEXT:    [[TMP133:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[CONV52:%.*]] = bitcast i64* [[N_CASTED51]] to i32*
// CHECK16-NEXT:    store i32 [[TMP133]], i32* [[CONV52]], align 4
// CHECK16-NEXT:    [[TMP134:%.*]] = load i64, i64* [[N_CASTED51]], align 8
// CHECK16-NEXT:    [[TMP135:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK16-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i64*
// CHECK16-NEXT:    store i64 [[TMP132]], i64* [[TMP137]], align 8
// CHECK16-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i64*
// CHECK16-NEXT:    store i64 [[TMP132]], i64* [[TMP139]], align 8
// CHECK16-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
// CHECK16-NEXT:    store i64 4, i64* [[TMP140]], align 8
// CHECK16-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP141]], align 8
// CHECK16-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i64*
// CHECK16-NEXT:    store i64 [[TMP134]], i64* [[TMP143]], align 8
// CHECK16-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP145:%.*]] = bitcast i8** [[TMP144]] to i64*
// CHECK16-NEXT:    store i64 [[TMP134]], i64* [[TMP145]], align 8
// CHECK16-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 1
// CHECK16-NEXT:    store i64 4, i64* [[TMP146]], align 8
// CHECK16-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 1
// CHECK16-NEXT:    store i8* null, i8** [[TMP147]], align 8
// CHECK16-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP149]], align 8
// CHECK16-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 2
// CHECK16-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
// CHECK16-NEXT:    store i64 [[TMP1]], i64* [[TMP151]], align 8
// CHECK16-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 2
// CHECK16-NEXT:    store i64 8, i64* [[TMP152]], align 8
// CHECK16-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 2
// CHECK16-NEXT:    store i8* null, i8** [[TMP153]], align 8
// CHECK16-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 3
// CHECK16-NEXT:    [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP155]], align 8
// CHECK16-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 3
// CHECK16-NEXT:    [[TMP157:%.*]] = bitcast i8** [[TMP156]] to i32**
// CHECK16-NEXT:    store i32* [[VLA]], i32** [[TMP157]], align 8
// CHECK16-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 3
// CHECK16-NEXT:    store i64 [[TMP135]], i64* [[TMP158]], align 8
// CHECK16-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 3
// CHECK16-NEXT:    store i8* null, i8** [[TMP159]], align 8
// CHECK16-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP163:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    store i32 [[TMP163]], i32* [[DOTCAPTURE_EXPR_58]], align 4
// CHECK16-NEXT:    [[TMP164:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_58]], align 4
// CHECK16-NEXT:    [[SUB60:%.*]] = sub nsw i32 [[TMP164]], 0
// CHECK16-NEXT:    [[DIV61:%.*]] = sdiv i32 [[SUB60]], 1
// CHECK16-NEXT:    [[SUB62:%.*]] = sub nsw i32 [[DIV61]], 1
// CHECK16-NEXT:    store i32 [[SUB62]], i32* [[DOTCAPTURE_EXPR_59]], align 4
// CHECK16-NEXT:    [[TMP165:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_59]], align 4
// CHECK16-NEXT:    [[ADD63:%.*]] = add nsw i32 [[TMP165]], 1
// CHECK16-NEXT:    [[TMP166:%.*]] = zext i32 [[ADD63]] to i64
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP166]])
// CHECK16-NEXT:    [[TMP167:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174.region_id, i32 4, i8** [[TMP160]], i8** [[TMP161]], i64* [[TMP162]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP168:%.*]] = icmp ne i32 [[TMP167]], 0
// CHECK16-NEXT:    br i1 [[TMP168]], label [[OMP_OFFLOAD_FAILED64:%.*]], label [[OMP_OFFLOAD_CONT65:%.*]]
// CHECK16:       omp_offload.failed64:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i64 [[TMP132]], i64 [[TMP134]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT65]]
// CHECK16:       omp_offload.cont65:
// CHECK16-NEXT:    [[TMP169:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK16-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP169]])
// CHECK16-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK16-NEXT:    [[TMP170:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK16-NEXT:    call void @llvm.stackrestore(i8* [[TMP170]])
// CHECK16-NEXT:    [[TMP171:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK16-NEXT:    ret i32 [[TMP171]]
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK16-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK16-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK16-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !13
// CHECK16-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK16-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !13
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !13
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK16-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK16-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK16-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK16-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK16-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK16-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !17
// CHECK16-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK16-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !17
// CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !17
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !17
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK16-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK16-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK16-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK16-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK16-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK16-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK16-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK16-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK16-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK16-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
// CHECK16-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK16-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !22
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK16-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK16-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK16-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK16-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK16-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK16-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
// CHECK16-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK16-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !25
// CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !25
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK16-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK16-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK16-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK16-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK16-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK16-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK16-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK16-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK16-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK16-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK16-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK16-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*, i64, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK16-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK16-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !28
// CHECK16-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP21]], i32* [[CONV7]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !28
// CHECK16-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, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]]), !llvm.access.group !28
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK16-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK16-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK16-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
// CHECK16-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK16:       cond.true12:
// CHECK16-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    br label [[COND_END14:%.*]]
// CHECK16:       cond.false13:
// CHECK16-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    br label [[COND_END14]]
// CHECK16:       cond.end14:
// CHECK16-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
// CHECK16-NEXT:    store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
// CHECK16-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK16-NEXT:    br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK16-NEXT:    [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
// CHECK16-NEXT:    [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD18]], i32* [[I4]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK16-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK16-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK16-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !31
// CHECK16-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK16-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !31
// CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !31
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !31
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK16-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK16-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK16-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK16-NEXT:    [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
// CHECK16-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
// CHECK16-NEXT:    [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
// CHECK16-NEXT:    store i32 [[ADD13]], i32* [[I6]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK16-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK16-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..8
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK16-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !34
// CHECK16-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK16-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !34
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK16-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK16-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK16-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK16-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK16-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK16-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK16-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK16:       omp.dispatch.cond:
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK16-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK16-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK16-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK16:       omp.dispatch.body:
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !37
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !37
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !37
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !37
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK16-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK16-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK16:       omp.dispatch.inc:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK16:       omp.dispatch.end:
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK16-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK16-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK16-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK16-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK16-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK16-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK16-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK16-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK16-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK16-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*, i64, i32*, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK16-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK16-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !40
// CHECK16-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK16-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !40
// CHECK16-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP20]], i32* [[CONV7]], align 4, !llvm.access.group !40
// CHECK16-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !40
// CHECK16-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, i32*, i64, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]]), !llvm.access.group !40
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !40
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK16-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
// CHECK16-NEXT:    br i1 [[TMP27]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK16-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK16-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK16-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK16-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK16:       omp.precond.then:
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK16-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK16-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK16-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK16-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK16:       omp.dispatch.cond:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK16-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK16-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK16-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK16:       omp.dispatch.body:
// CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
// CHECK16-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK16-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !43
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !43
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !43
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK16-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK16-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK16:       omp.dispatch.inc:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK16:       omp.dispatch.end:
// CHECK16-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK16-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK16-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
// CHECK16-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
// CHECK16-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK16-NEXT:    store i32 [[ADD12]], i32* [[I6]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK16:       omp.precond.end:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK16-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK16-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
// CHECK16-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 8
// CHECK16-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[M_CASTED19:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [2 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_PTRS22:%.*]] = alloca [2 x i8*], align 8
// CHECK16-NEXT:    [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [2 x i8*], align 8
// CHECK16-NEXT:    [[_TMP24:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK16-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK16-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
// CHECK16-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK16-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK16-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK16-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK16:       omp_offload.failed:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK16:       omp_offload.cont:
// CHECK16-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
// CHECK16-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
// CHECK16-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK16-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK16-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK16-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK16:       omp_offload.failed5:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK16:       omp_offload.cont6:
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[M]], align 4
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP18]], i32* [[CONV]], align 4
// CHECK16-NEXT:    [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK16-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
// CHECK16-NEXT:    store i64 [[TMP19]], i64* [[TMP21]], align 8
// CHECK16-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK16-NEXT:    store i64 [[TMP19]], i64* [[TMP23]], align 8
// CHECK16-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP24]], align 8
// CHECK16-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
// CHECK16-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
// CHECK16-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
// CHECK16-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK16-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK16-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
// CHECK16-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
// CHECK16:       omp_offload.failed11:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
// CHECK16:       omp_offload.cont12:
// CHECK16-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 8
// CHECK16-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
// CHECK16-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK16-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK16-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK16-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK16:       omp_offload.failed17:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK16:       omp_offload.cont18:
// CHECK16-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK16-NEXT:    [[CONV20:%.*]] = bitcast i64* [[M_CASTED19]] to i32*
// CHECK16-NEXT:    store i32 [[TMP43]], i32* [[CONV20]], align 4
// CHECK16-NEXT:    [[TMP44:%.*]] = load i64, i64* [[M_CASTED19]], align 8
// CHECK16-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK16-NEXT:    store i64 [[TMP44]], i64* [[TMP46]], align 8
// CHECK16-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK16-NEXT:    store i64 [[TMP44]], i64* [[TMP48]], align 8
// CHECK16-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
// CHECK16-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK16-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 8
// CHECK16-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
// CHECK16-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 8
// CHECK16-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
// CHECK16-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK16-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
// CHECK16-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK16-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK16-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
// CHECK16-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED25:%.*]], label [[OMP_OFFLOAD_CONT26:%.*]]
// CHECK16:       omp_offload.failed25:
// CHECK16-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i64 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK16-NEXT:    br label [[OMP_OFFLOAD_CONT26]]
// CHECK16:       omp_offload.cont26:
// CHECK16-NEXT:    ret i32 0
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK16-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK16-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK16-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !46
// CHECK16-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK16-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !46
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !46
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP47:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK16-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK16-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK16-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !49
// CHECK16-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK16-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !49
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !49
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !49
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK16-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK16-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP50:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK16-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK16-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK16-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK16-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !52
// CHECK16-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK16-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !52
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !52
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK16-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK16-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK16-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !55
// CHECK16-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK16-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !55
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !55
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !55
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK16-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK16-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK16-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK16-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK16-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK16-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK16-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !58
// CHECK16-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK16-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !58
// CHECK16-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !58
// CHECK16-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !58
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !58
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !58
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK16-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK16-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK16-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK16:       omp.dispatch.cond:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK16-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK16:       omp.dispatch.body:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
// CHECK16-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK16-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK16-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK16-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK16:       omp.dispatch.inc:
// CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK16-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK16-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK16:       omp.dispatch.end:
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK16-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK16-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK16-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK16-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK16-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !64
// CHECK16-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK16-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !64
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !64
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP65:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK16-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK16-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK16-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK16-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK16:       omp.dispatch.cond:
// CHECK16-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK16-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK16-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK16:       omp.dispatch.body:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !67
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !67
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !67
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !67
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK16-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK16-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP68:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK16:       omp.dispatch.inc:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK16:       omp.dispatch.end:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK16-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK16-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK16-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK16:       cond.true:
// CHECK16-NEXT:    br label [[COND_END:%.*]]
// CHECK16:       cond.false:
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    br label [[COND_END]]
// CHECK16:       cond.end:
// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK16-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK16-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !70
// CHECK16-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK16-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !70
// CHECK16-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK16-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !70
// CHECK16-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !70
// CHECK16-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !70
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !70
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP71:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK16:       omp.loop.exit:
// CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK16-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK16-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK16-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK16-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK16-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK16-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK16:       omp.dispatch.cond:
// CHECK16-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK16-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK16-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK16:       omp.dispatch.body:
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK16-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK16:       omp.inner.for.cond:
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !73
// CHECK16-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK16:       omp.inner.for.body:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !73
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !73
// CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !73
// CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK16:       omp.body.continue:
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK16:       omp.inner.for.inc:
// CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK16-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK16-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP74:![0-9]+]]
// CHECK16:       omp.inner.for.end:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK16:       omp.dispatch.inc:
// CHECK16-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK16:       omp.dispatch.end:
// CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK16-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK16-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK16:       .omp.final.then:
// CHECK16-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK16:       .omp.final.done:
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK16-SAME: () #[[ATTR5:[0-9]+]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK16-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@main
// CHECK17-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 4
// CHECK17-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
// CHECK17-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED17:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
// CHECK17-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED31:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS32:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS33:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS34:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES35:%.*]] = alloca [3 x i64], align 4
// CHECK17-NEXT:    [[_TMP36:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_37:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED45:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED46:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS47:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS48:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS49:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES50:%.*]] = alloca [4 x i64], align 4
// CHECK17-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_52:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_53:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK17-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK17-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK17-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK17-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK17-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK17-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[N_CASTED]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
// CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[TMP7]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[TMP9]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    store i64 4, i64* [[TMP10]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP13]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP15]], align 4
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK17-NEXT:    store i64 4, i64* [[TMP16]], align 4
// CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP19]], align 4
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP21]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP27:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK17-NEXT:    [[TMP30:%.*]] = zext i32 [[ADD]] to i64
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
// CHECK17-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK17-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK17:       omp_offload.failed:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
// CHECK17-NEXT:    [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
// CHECK17-NEXT:    [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
// CHECK17-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
// CHECK17-NEXT:    store i32 [[TMP34]], i32* [[TMP38]], align 4
// CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
// CHECK17-NEXT:    store i32 [[TMP34]], i32* [[TMP40]], align 4
// CHECK17-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK17-NEXT:    store i64 4, i64* [[TMP41]], align 4
// CHECK17-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP42]], align 4
// CHECK17-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP44]], align 4
// CHECK17-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP46]], align 4
// CHECK17-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
// CHECK17-NEXT:    store i64 4, i64* [[TMP47]], align 4
// CHECK17-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP48]], align 4
// CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP50]], align 4
// CHECK17-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP52]], align 4
// CHECK17-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP36]], i64* [[TMP53]], align 4
// CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP58:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK17-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK17-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
// CHECK17-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
// CHECK17-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
// CHECK17-NEXT:    store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK17-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK17-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
// CHECK17-NEXT:    [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
// CHECK17-NEXT:    [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159.region_id, i32 3, i8** [[TMP55]], i8** [[TMP56]], i64* [[TMP57]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
// CHECK17-NEXT:    br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
// CHECK17:       omp_offload.failed15:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT16]]
// CHECK17:       omp_offload.cont16:
// CHECK17-NEXT:    [[TMP64:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    store i32 [[TMP64]], i32* [[M_CASTED]], align 4
// CHECK17-NEXT:    [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK17-NEXT:    [[TMP66:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
// CHECK17-NEXT:    [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
// CHECK17-NEXT:    [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
// CHECK17-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
// CHECK17-NEXT:    store i32 [[TMP65]], i32* [[TMP71]], align 4
// CHECK17-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
// CHECK17-NEXT:    store i32 [[TMP65]], i32* [[TMP73]], align 4
// CHECK17-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
// CHECK17-NEXT:    store i64 4, i64* [[TMP74]], align 4
// CHECK17-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP75]], align 4
// CHECK17-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
// CHECK17-NEXT:    store i32 [[TMP67]], i32* [[TMP77]], align 4
// CHECK17-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
// CHECK17-NEXT:    store i32 [[TMP67]], i32* [[TMP79]], align 4
// CHECK17-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
// CHECK17-NEXT:    store i64 4, i64* [[TMP80]], align 4
// CHECK17-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP81]], align 4
// CHECK17-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP83]], align 4
// CHECK17-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP85]], align 4
// CHECK17-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
// CHECK17-NEXT:    store i64 4, i64* [[TMP86]], align 4
// CHECK17-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP87]], align 4
// CHECK17-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP89]], align 4
// CHECK17-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP91]], align 4
// CHECK17-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP69]], i64* [[TMP92]], align 4
// CHECK17-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP93]], align 4
// CHECK17-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP97:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK17-NEXT:    [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK17-NEXT:    [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
// CHECK17-NEXT:    [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
// CHECK17-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
// CHECK17-NEXT:    store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK17-NEXT:    [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK17-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
// CHECK17-NEXT:    [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
// CHECK17-NEXT:    [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164.region_id, i32 4, i8** [[TMP94]], i8** [[TMP95]], i64* [[TMP96]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
// CHECK17-NEXT:    br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
// CHECK17:       omp_offload.failed29:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT30]]
// CHECK17:       omp_offload.cont30:
// CHECK17-NEXT:    [[TMP103:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP103]], i32* [[N_CASTED31]], align 4
// CHECK17-NEXT:    [[TMP104:%.*]] = load i32, i32* [[N_CASTED31]], align 4
// CHECK17-NEXT:    [[TMP105:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP106:%.*]] = sext i32 [[TMP105]] to i64
// CHECK17-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP108:%.*]] = bitcast i8** [[TMP107]] to i32*
// CHECK17-NEXT:    store i32 [[TMP104]], i32* [[TMP108]], align 4
// CHECK17-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
// CHECK17-NEXT:    store i32 [[TMP104]], i32* [[TMP110]], align 4
// CHECK17-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
// CHECK17-NEXT:    store i64 4, i64* [[TMP111]], align 4
// CHECK17-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP112]], align 4
// CHECK17-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP114]], align 4
// CHECK17-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP116:%.*]] = bitcast i8** [[TMP115]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP116]], align 4
// CHECK17-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 1
// CHECK17-NEXT:    store i64 4, i64* [[TMP117]], align 4
// CHECK17-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP118]], align 4
// CHECK17-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP120]], align 4
// CHECK17-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP122]], align 4
// CHECK17-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP106]], i64* [[TMP123]], align 4
// CHECK17-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP124]], align 4
// CHECK17-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP128:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP128]], i32* [[DOTCAPTURE_EXPR_37]], align 4
// CHECK17-NEXT:    [[TMP129:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_37]], align 4
// CHECK17-NEXT:    [[SUB39:%.*]] = sub nsw i32 [[TMP129]], 0
// CHECK17-NEXT:    [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
// CHECK17-NEXT:    [[SUB41:%.*]] = sub nsw i32 [[DIV40]], 1
// CHECK17-NEXT:    store i32 [[SUB41]], i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK17-NEXT:    [[TMP130:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK17-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP130]], 1
// CHECK17-NEXT:    [[TMP131:%.*]] = zext i32 [[ADD42]] to i64
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP131]])
// CHECK17-NEXT:    [[TMP132:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169.region_id, i32 3, i8** [[TMP125]], i8** [[TMP126]], i64* [[TMP127]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP133:%.*]] = icmp ne i32 [[TMP132]], 0
// CHECK17-NEXT:    br i1 [[TMP133]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
// CHECK17:       omp_offload.failed43:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i32 [[TMP104]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT44]]
// CHECK17:       omp_offload.cont44:
// CHECK17-NEXT:    [[TMP134:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    store i32 [[TMP134]], i32* [[M_CASTED45]], align 4
// CHECK17-NEXT:    [[TMP135:%.*]] = load i32, i32* [[M_CASTED45]], align 4
// CHECK17-NEXT:    [[TMP136:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP136]], i32* [[N_CASTED46]], align 4
// CHECK17-NEXT:    [[TMP137:%.*]] = load i32, i32* [[N_CASTED46]], align 4
// CHECK17-NEXT:    [[TMP138:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP139:%.*]] = sext i32 [[TMP138]] to i64
// CHECK17-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32*
// CHECK17-NEXT:    store i32 [[TMP135]], i32* [[TMP141]], align 4
// CHECK17-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i32*
// CHECK17-NEXT:    store i32 [[TMP135]], i32* [[TMP143]], align 4
// CHECK17-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
// CHECK17-NEXT:    store i64 4, i64* [[TMP144]], align 4
// CHECK17-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP145]], align 4
// CHECK17-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
// CHECK17-NEXT:    store i32 [[TMP137]], i32* [[TMP147]], align 4
// CHECK17-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
// CHECK17-NEXT:    store i32 [[TMP137]], i32* [[TMP149]], align 4
// CHECK17-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 1
// CHECK17-NEXT:    store i64 4, i64* [[TMP150]], align 4
// CHECK17-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP151]], align 4
// CHECK17-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP153]], align 4
// CHECK17-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP155]], align 4
// CHECK17-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 2
// CHECK17-NEXT:    store i64 4, i64* [[TMP156]], align 4
// CHECK17-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP157]], align 4
// CHECK17-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP159:%.*]] = bitcast i8** [[TMP158]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP159]], align 4
// CHECK17-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP161:%.*]] = bitcast i8** [[TMP160]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP161]], align 4
// CHECK17-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP139]], i64* [[TMP162]], align 4
// CHECK17-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP163]], align 4
// CHECK17-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP167:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP167]], i32* [[DOTCAPTURE_EXPR_52]], align 4
// CHECK17-NEXT:    [[TMP168:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_52]], align 4
// CHECK17-NEXT:    [[SUB54:%.*]] = sub nsw i32 [[TMP168]], 0
// CHECK17-NEXT:    [[DIV55:%.*]] = sdiv i32 [[SUB54]], 1
// CHECK17-NEXT:    [[SUB56:%.*]] = sub nsw i32 [[DIV55]], 1
// CHECK17-NEXT:    store i32 [[SUB56]], i32* [[DOTCAPTURE_EXPR_53]], align 4
// CHECK17-NEXT:    [[TMP169:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_53]], align 4
// CHECK17-NEXT:    [[ADD57:%.*]] = add nsw i32 [[TMP169]], 1
// CHECK17-NEXT:    [[TMP170:%.*]] = zext i32 [[ADD57]] to i64
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP170]])
// CHECK17-NEXT:    [[TMP171:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174.region_id, i32 4, i8** [[TMP164]], i8** [[TMP165]], i64* [[TMP166]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP172:%.*]] = icmp ne i32 [[TMP171]], 0
// CHECK17-NEXT:    br i1 [[TMP172]], label [[OMP_OFFLOAD_FAILED58:%.*]], label [[OMP_OFFLOAD_CONT59:%.*]]
// CHECK17:       omp_offload.failed58:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i32 [[TMP135]], i32 [[TMP137]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT59]]
// CHECK17:       omp_offload.cont59:
// CHECK17-NEXT:    [[TMP173:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK17-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP173]])
// CHECK17-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK17-NEXT:    [[TMP174:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP174]])
// CHECK17-NEXT:    [[TMP175:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK17-NEXT:    ret i32 [[TMP175]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK17-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !14
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK17-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK17-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK17-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK17-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK17-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !23
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK17-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK17-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK17-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK17-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK17-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-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*, i32, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK17-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK17-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, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]]), !llvm.access.group !29
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
// CHECK17-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK17:       cond.true11:
// CHECK17-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    br label [[COND_END13:%.*]]
// CHECK17:       cond.false12:
// CHECK17-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    br label [[COND_END13]]
// CHECK17:       cond.end13:
// CHECK17-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
// CHECK17-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
// CHECK17-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK17-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK17-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
// CHECK17-NEXT:    [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD17]], i32* [[I4]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK17-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK17-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK17-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK17-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK17-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK17-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..8
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !35
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK17-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK17-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK17-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK17-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK17-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK17-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK17-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK17-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-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*, i32, i32*, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK17-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, i32*, i32, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]]), !llvm.access.group !41
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK17-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK17-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[I4]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK17-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK17-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK17-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK17-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK17-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK17-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK17-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK17-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK17-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK17-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED19:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK17-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK17-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK17-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK17:       omp_offload.failed:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK17-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK17-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK17:       omp_offload.failed5:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK17:       omp_offload.cont6:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    store i32 [[TMP18]], i32* [[M_CASTED]], align 4
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
// CHECK17-NEXT:    store i32 [[TMP19]], i32* [[TMP21]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK17-NEXT:    store i32 [[TMP19]], i32* [[TMP23]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP24]], align 4
// CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
// CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
// CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP29]], align 4
// CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK17-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
// CHECK17-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
// CHECK17:       omp_offload.failed11:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
// CHECK17:       omp_offload.cont12:
// CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 4
// CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
// CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP38]], align 4
// CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK17-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK17-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK17:       omp_offload.failed17:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK17:       omp_offload.cont18:
// CHECK17-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    store i32 [[TMP43]], i32* [[M_CASTED19]], align 4
// CHECK17-NEXT:    [[TMP44:%.*]] = load i32, i32* [[M_CASTED19]], align 4
// CHECK17-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK17-NEXT:    store i32 [[TMP44]], i32* [[TMP46]], align 4
// CHECK17-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
// CHECK17-NEXT:    store i32 [[TMP44]], i32* [[TMP48]], align 4
// CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP49]], align 4
// CHECK17-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 4
// CHECK17-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 4
// CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
// CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK17-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK17-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
// CHECK17-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK17:       omp_offload.failed24:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i32 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK17:       omp_offload.cont25:
// CHECK17-NEXT:    ret i32 0
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK17-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !47
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK17-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK17-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK17-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !53
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK17-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK17-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK17-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !59
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK17-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK17-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK17-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK17-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK17-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !65
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP66:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK17-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK17-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK17-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK17-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK17-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP69:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK17-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK17-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !71
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP72:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK17-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK17-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK17-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK17-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK17-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP75:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK17-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK17-SAME: () #[[ATTR5:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK17-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@main
// CHECK18-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 4
// CHECK18-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_CASTED3:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
// CHECK18-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_CASTED17:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
// CHECK18-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_CASTED31:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS32:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS33:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS34:%.*]] = alloca [3 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_SIZES35:%.*]] = alloca [3 x i64], align 4
// CHECK18-NEXT:    [[_TMP36:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_37:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[M_CASTED45:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_CASTED46:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS47:%.*]] = alloca [4 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS48:%.*]] = alloca [4 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS49:%.*]] = alloca [4 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_SIZES50:%.*]] = alloca [4 x i64], align 4
// CHECK18-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_52:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_53:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK18-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK18-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK18-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK18-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK18-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK18-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[N_CASTED]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK18-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
// CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[TMP7]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[TMP9]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK18-NEXT:    store i64 4, i64* [[TMP10]], align 4
// CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP13]], align 4
// CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP15]], align 4
// CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK18-NEXT:    store i64 4, i64* [[TMP16]], align 4
// CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP19]], align 4
// CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP21]], align 4
// CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 4
// CHECK18-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP27:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK18-NEXT:    [[TMP30:%.*]] = zext i32 [[ADD]] to i64
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
// CHECK18-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK18-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK18:       omp_offload.failed:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
// CHECK18-NEXT:    [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
// CHECK18-NEXT:    [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK18-NEXT:    [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
// CHECK18-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
// CHECK18-NEXT:    store i32 [[TMP34]], i32* [[TMP38]], align 4
// CHECK18-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
// CHECK18-NEXT:    store i32 [[TMP34]], i32* [[TMP40]], align 4
// CHECK18-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK18-NEXT:    store i64 4, i64* [[TMP41]], align 4
// CHECK18-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP42]], align 4
// CHECK18-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP44]], align 4
// CHECK18-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP46]], align 4
// CHECK18-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
// CHECK18-NEXT:    store i64 4, i64* [[TMP47]], align 4
// CHECK18-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP48]], align 4
// CHECK18-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP50]], align 4
// CHECK18-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP52]], align 4
// CHECK18-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
// CHECK18-NEXT:    store i64 [[TMP36]], i64* [[TMP53]], align 4
// CHECK18-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK18-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP58:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK18-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK18-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
// CHECK18-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
// CHECK18-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
// CHECK18-NEXT:    store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK18-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK18-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
// CHECK18-NEXT:    [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
// CHECK18-NEXT:    [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159.region_id, i32 3, i8** [[TMP55]], i8** [[TMP56]], i64* [[TMP57]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
// CHECK18-NEXT:    br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
// CHECK18:       omp_offload.failed15:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT16]]
// CHECK18:       omp_offload.cont16:
// CHECK18-NEXT:    [[TMP64:%.*]] = load i32, i32* [[M]], align 4
// CHECK18-NEXT:    store i32 [[TMP64]], i32* [[M_CASTED]], align 4
// CHECK18-NEXT:    [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK18-NEXT:    [[TMP66:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
// CHECK18-NEXT:    [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
// CHECK18-NEXT:    [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK18-NEXT:    [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
// CHECK18-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
// CHECK18-NEXT:    store i32 [[TMP65]], i32* [[TMP71]], align 4
// CHECK18-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
// CHECK18-NEXT:    store i32 [[TMP65]], i32* [[TMP73]], align 4
// CHECK18-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
// CHECK18-NEXT:    store i64 4, i64* [[TMP74]], align 4
// CHECK18-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP75]], align 4
// CHECK18-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
// CHECK18-NEXT:    store i32 [[TMP67]], i32* [[TMP77]], align 4
// CHECK18-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
// CHECK18-NEXT:    store i32 [[TMP67]], i32* [[TMP79]], align 4
// CHECK18-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
// CHECK18-NEXT:    store i64 4, i64* [[TMP80]], align 4
// CHECK18-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP81]], align 4
// CHECK18-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP83]], align 4
// CHECK18-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP85]], align 4
// CHECK18-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
// CHECK18-NEXT:    store i64 4, i64* [[TMP86]], align 4
// CHECK18-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP87]], align 4
// CHECK18-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP89]], align 4
// CHECK18-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP91]], align 4
// CHECK18-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
// CHECK18-NEXT:    store i64 [[TMP69]], i64* [[TMP92]], align 4
// CHECK18-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
// CHECK18-NEXT:    store i8* null, i8** [[TMP93]], align 4
// CHECK18-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP97:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK18-NEXT:    [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK18-NEXT:    [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
// CHECK18-NEXT:    [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
// CHECK18-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
// CHECK18-NEXT:    store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK18-NEXT:    [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK18-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
// CHECK18-NEXT:    [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
// CHECK18-NEXT:    [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164.region_id, i32 4, i8** [[TMP94]], i8** [[TMP95]], i64* [[TMP96]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
// CHECK18-NEXT:    br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
// CHECK18:       omp_offload.failed29:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT30]]
// CHECK18:       omp_offload.cont30:
// CHECK18-NEXT:    [[TMP103:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP103]], i32* [[N_CASTED31]], align 4
// CHECK18-NEXT:    [[TMP104:%.*]] = load i32, i32* [[N_CASTED31]], align 4
// CHECK18-NEXT:    [[TMP105:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK18-NEXT:    [[TMP106:%.*]] = sext i32 [[TMP105]] to i64
// CHECK18-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP108:%.*]] = bitcast i8** [[TMP107]] to i32*
// CHECK18-NEXT:    store i32 [[TMP104]], i32* [[TMP108]], align 4
// CHECK18-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
// CHECK18-NEXT:    store i32 [[TMP104]], i32* [[TMP110]], align 4
// CHECK18-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
// CHECK18-NEXT:    store i64 4, i64* [[TMP111]], align 4
// CHECK18-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP112]], align 4
// CHECK18-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP114]], align 4
// CHECK18-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP116:%.*]] = bitcast i8** [[TMP115]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP116]], align 4
// CHECK18-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 1
// CHECK18-NEXT:    store i64 4, i64* [[TMP117]], align 4
// CHECK18-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP118]], align 4
// CHECK18-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP120]], align 4
// CHECK18-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP122]], align 4
// CHECK18-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 2
// CHECK18-NEXT:    store i64 [[TMP106]], i64* [[TMP123]], align 4
// CHECK18-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP124]], align 4
// CHECK18-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP128:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP128]], i32* [[DOTCAPTURE_EXPR_37]], align 4
// CHECK18-NEXT:    [[TMP129:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_37]], align 4
// CHECK18-NEXT:    [[SUB39:%.*]] = sub nsw i32 [[TMP129]], 0
// CHECK18-NEXT:    [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
// CHECK18-NEXT:    [[SUB41:%.*]] = sub nsw i32 [[DIV40]], 1
// CHECK18-NEXT:    store i32 [[SUB41]], i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK18-NEXT:    [[TMP130:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK18-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP130]], 1
// CHECK18-NEXT:    [[TMP131:%.*]] = zext i32 [[ADD42]] to i64
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP131]])
// CHECK18-NEXT:    [[TMP132:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169.region_id, i32 3, i8** [[TMP125]], i8** [[TMP126]], i64* [[TMP127]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP133:%.*]] = icmp ne i32 [[TMP132]], 0
// CHECK18-NEXT:    br i1 [[TMP133]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
// CHECK18:       omp_offload.failed43:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i32 [[TMP104]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT44]]
// CHECK18:       omp_offload.cont44:
// CHECK18-NEXT:    [[TMP134:%.*]] = load i32, i32* [[M]], align 4
// CHECK18-NEXT:    store i32 [[TMP134]], i32* [[M_CASTED45]], align 4
// CHECK18-NEXT:    [[TMP135:%.*]] = load i32, i32* [[M_CASTED45]], align 4
// CHECK18-NEXT:    [[TMP136:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP136]], i32* [[N_CASTED46]], align 4
// CHECK18-NEXT:    [[TMP137:%.*]] = load i32, i32* [[N_CASTED46]], align 4
// CHECK18-NEXT:    [[TMP138:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK18-NEXT:    [[TMP139:%.*]] = sext i32 [[TMP138]] to i64
// CHECK18-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32*
// CHECK18-NEXT:    store i32 [[TMP135]], i32* [[TMP141]], align 4
// CHECK18-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i32*
// CHECK18-NEXT:    store i32 [[TMP135]], i32* [[TMP143]], align 4
// CHECK18-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
// CHECK18-NEXT:    store i64 4, i64* [[TMP144]], align 4
// CHECK18-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP145]], align 4
// CHECK18-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
// CHECK18-NEXT:    store i32 [[TMP137]], i32* [[TMP147]], align 4
// CHECK18-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
// CHECK18-NEXT:    store i32 [[TMP137]], i32* [[TMP149]], align 4
// CHECK18-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 1
// CHECK18-NEXT:    store i64 4, i64* [[TMP150]], align 4
// CHECK18-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP151]], align 4
// CHECK18-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP153]], align 4
// CHECK18-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[TMP155]], align 4
// CHECK18-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 2
// CHECK18-NEXT:    store i64 4, i64* [[TMP156]], align 4
// CHECK18-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP157]], align 4
// CHECK18-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP159:%.*]] = bitcast i8** [[TMP158]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP159]], align 4
// CHECK18-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP161:%.*]] = bitcast i8** [[TMP160]] to i32**
// CHECK18-NEXT:    store i32* [[VLA]], i32** [[TMP161]], align 4
// CHECK18-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 3
// CHECK18-NEXT:    store i64 [[TMP139]], i64* [[TMP162]], align 4
// CHECK18-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 3
// CHECK18-NEXT:    store i8* null, i8** [[TMP163]], align 4
// CHECK18-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP167:%.*]] = load i32, i32* [[N]], align 4
// CHECK18-NEXT:    store i32 [[TMP167]], i32* [[DOTCAPTURE_EXPR_52]], align 4
// CHECK18-NEXT:    [[TMP168:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_52]], align 4
// CHECK18-NEXT:    [[SUB54:%.*]] = sub nsw i32 [[TMP168]], 0
// CHECK18-NEXT:    [[DIV55:%.*]] = sdiv i32 [[SUB54]], 1
// CHECK18-NEXT:    [[SUB56:%.*]] = sub nsw i32 [[DIV55]], 1
// CHECK18-NEXT:    store i32 [[SUB56]], i32* [[DOTCAPTURE_EXPR_53]], align 4
// CHECK18-NEXT:    [[TMP169:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_53]], align 4
// CHECK18-NEXT:    [[ADD57:%.*]] = add nsw i32 [[TMP169]], 1
// CHECK18-NEXT:    [[TMP170:%.*]] = zext i32 [[ADD57]] to i64
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP170]])
// CHECK18-NEXT:    [[TMP171:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174.region_id, i32 4, i8** [[TMP164]], i8** [[TMP165]], i64* [[TMP166]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP172:%.*]] = icmp ne i32 [[TMP171]], 0
// CHECK18-NEXT:    br i1 [[TMP172]], label [[OMP_OFFLOAD_FAILED58:%.*]], label [[OMP_OFFLOAD_CONT59:%.*]]
// CHECK18:       omp_offload.failed58:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i32 [[TMP135]], i32 [[TMP137]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT59]]
// CHECK18:       omp_offload.cont59:
// CHECK18-NEXT:    [[TMP173:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK18-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP173]])
// CHECK18-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK18-NEXT:    [[TMP174:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP174]])
// CHECK18-NEXT:    [[TMP175:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK18-NEXT:    ret i32 [[TMP175]]
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK18-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK18-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !14
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !14
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !14
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK18-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK18-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK18-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !18
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !18
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK18-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK18-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK18-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK18-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK18-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK18-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK18-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK18-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK18-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !23
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK18-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK18-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK18-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !26
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !26
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK18-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK18-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK18-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK18-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK18-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK18-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK18-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK18-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK18-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*, i32, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK18-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK18-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, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]]), !llvm.access.group !29
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK18-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
// CHECK18-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK18:       cond.true11:
// CHECK18-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    br label [[COND_END13:%.*]]
// CHECK18:       cond.false12:
// CHECK18-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    br label [[COND_END13]]
// CHECK18:       cond.end13:
// CHECK18-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
// CHECK18-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
// CHECK18-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK18-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK18-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
// CHECK18-NEXT:    [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD17]], i32* [[I4]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !32
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !32
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK18-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK18-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK18-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK18-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK18-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK18-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK18-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..8
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK18-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !35
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK18-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK18-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK18-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK18:       omp.dispatch.cond:
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK18-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK18-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK18:       omp.dispatch.body:
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK18-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK18-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !38
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !38
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK18-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK18-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK18:       omp.dispatch.inc:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK18:       omp.dispatch.end:
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK18-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK18-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK18-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK18-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK18-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*, i32, i32*, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK18-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, i32*, i32, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]]), !llvm.access.group !41
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK18-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK18-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[I4]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK18-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK18-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK18-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK18:       omp.precond.then:
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK18-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK18-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK18:       omp.dispatch.cond:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK18-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK18-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK18-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK18:       omp.dispatch.body:
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !44
// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !44
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !44
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !44
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK18-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK18-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK18:       omp.dispatch.inc:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK18:       omp.dispatch.end:
// CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK18-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK18-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK18-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK18-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK18-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK18:       omp.precond.end:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK18-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK18-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
// CHECK18-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 4
// CHECK18-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[M_CASTED19:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [2 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [2 x i8*], align 4
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [2 x i8*], align 4
// CHECK18-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK18-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK18-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK18-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK18:       omp_offload.failed:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
// CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK18-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK18-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK18:       omp_offload.failed5:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK18:       omp_offload.cont6:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[M]], align 4
// CHECK18-NEXT:    store i32 [[TMP18]], i32* [[M_CASTED]], align 4
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
// CHECK18-NEXT:    store i32 [[TMP19]], i32* [[TMP21]], align 4
// CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK18-NEXT:    store i32 [[TMP19]], i32* [[TMP23]], align 4
// CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP24]], align 4
// CHECK18-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
// CHECK18-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
// CHECK18-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP29]], align 4
// CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK18-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
// CHECK18-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
// CHECK18:       omp_offload.failed11:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
// CHECK18:       omp_offload.cont12:
// CHECK18-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 4
// CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
// CHECK18-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP38]], align 4
// CHECK18-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK18-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK18-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK18:       omp_offload.failed17:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK18:       omp_offload.cont18:
// CHECK18-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK18-NEXT:    store i32 [[TMP43]], i32* [[M_CASTED19]], align 4
// CHECK18-NEXT:    [[TMP44:%.*]] = load i32, i32* [[M_CASTED19]], align 4
// CHECK18-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK18-NEXT:    store i32 [[TMP44]], i32* [[TMP46]], align 4
// CHECK18-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
// CHECK18-NEXT:    store i32 [[TMP44]], i32* [[TMP48]], align 4
// CHECK18-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP49]], align 4
// CHECK18-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 4
// CHECK18-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 4
// CHECK18-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK18-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
// CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK18-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK18-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
// CHECK18-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK18:       omp_offload.failed24:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i32 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK18:       omp_offload.cont25:
// CHECK18-NEXT:    ret i32 0
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK18-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !47
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK18-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
// CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !50
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !50
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK18-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK18-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK18-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !53
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !53
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !53
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK18-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !56
// CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !56
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !56
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !56
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK18-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK18-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK18-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !59
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK18-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK18:       omp.dispatch.cond:
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK18:       omp.dispatch.body:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
// CHECK18-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK18-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
// CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !62
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !62
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK18-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK18:       omp.dispatch.inc:
// CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK18-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK18-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK18:       omp.dispatch.end:
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK18-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK18-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !65
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !65
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !65
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP66:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK18-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK18-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK18:       omp.dispatch.cond:
// CHECK18-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK18-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK18-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK18:       omp.dispatch.body:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !68
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !68
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !68
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !68
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK18-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK18-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP69:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK18:       omp.dispatch.inc:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK18:       omp.dispatch.end:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK18-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK18-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK18:       cond.true:
// CHECK18-NEXT:    br label [[COND_END:%.*]]
// CHECK18:       cond.false:
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    br label [[COND_END]]
// CHECK18:       cond.end:
// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !71
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP72:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK18:       omp.loop.exit:
// CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK18-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK18-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK18:       omp.dispatch.cond:
// CHECK18-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK18-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK18-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK18:       omp.dispatch.body:
// CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK18-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK18:       omp.inner.for.cond:
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !74
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK18:       omp.inner.for.body:
// CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !74
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !74
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
// CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !74
// CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK18:       omp.body.continue:
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK18:       omp.inner.for.inc:
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK18-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK18-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP75:![0-9]+]]
// CHECK18:       omp.inner.for.end:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK18:       omp.dispatch.inc:
// CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK18:       omp.dispatch.end:
// CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK18-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK18-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK18:       .omp.final.then:
// CHECK18-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK18-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK18:       .omp.final.done:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK18-SAME: () #[[ATTR5:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK18-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@main
// CHECK19-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 4
// CHECK19-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
// CHECK19-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED17:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
// CHECK19-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED31:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS32:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS33:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS34:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES35:%.*]] = alloca [3 x i64], align 4
// CHECK19-NEXT:    [[_TMP36:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_37:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[M_CASTED45:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED46:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS47:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS48:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS49:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES50:%.*]] = alloca [4 x i64], align 4
// CHECK19-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_52:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_53:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK19-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK19-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK19-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK19-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK19-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK19-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[N_CASTED]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
// CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP7]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP9]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    store i64 4, i64* [[TMP10]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP15]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK19-NEXT:    store i64 4, i64* [[TMP16]], align 4
// CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP19]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP21]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK19-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP27:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK19-NEXT:    [[TMP30:%.*]] = zext i32 [[ADD]] to i64
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
// CHECK19-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK19-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK19:       omp_offload.failed:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
// CHECK19-NEXT:    [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
// CHECK19-NEXT:    [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
// CHECK19-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
// CHECK19-NEXT:    store i32 [[TMP34]], i32* [[TMP38]], align 4
// CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
// CHECK19-NEXT:    store i32 [[TMP34]], i32* [[TMP40]], align 4
// CHECK19-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK19-NEXT:    store i64 4, i64* [[TMP41]], align 4
// CHECK19-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP42]], align 4
// CHECK19-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP44]], align 4
// CHECK19-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP46]], align 4
// CHECK19-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
// CHECK19-NEXT:    store i64 4, i64* [[TMP47]], align 4
// CHECK19-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP48]], align 4
// CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP50]], align 4
// CHECK19-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP52]], align 4
// CHECK19-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
// CHECK19-NEXT:    store i64 [[TMP36]], i64* [[TMP53]], align 4
// CHECK19-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK19-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP58:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK19-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK19-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
// CHECK19-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
// CHECK19-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
// CHECK19-NEXT:    store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK19-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK19-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
// CHECK19-NEXT:    [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
// CHECK19-NEXT:    [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159.region_id, i32 3, i8** [[TMP55]], i8** [[TMP56]], i64* [[TMP57]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
// CHECK19-NEXT:    br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
// CHECK19:       omp_offload.failed15:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT16]]
// CHECK19:       omp_offload.cont16:
// CHECK19-NEXT:    [[TMP64:%.*]] = load i32, i32* [[M]], align 4
// CHECK19-NEXT:    store i32 [[TMP64]], i32* [[M_CASTED]], align 4
// CHECK19-NEXT:    [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK19-NEXT:    [[TMP66:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
// CHECK19-NEXT:    [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
// CHECK19-NEXT:    [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
// CHECK19-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
// CHECK19-NEXT:    store i32 [[TMP65]], i32* [[TMP71]], align 4
// CHECK19-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
// CHECK19-NEXT:    store i32 [[TMP65]], i32* [[TMP73]], align 4
// CHECK19-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
// CHECK19-NEXT:    store i64 4, i64* [[TMP74]], align 4
// CHECK19-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP75]], align 4
// CHECK19-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
// CHECK19-NEXT:    store i32 [[TMP67]], i32* [[TMP77]], align 4
// CHECK19-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
// CHECK19-NEXT:    store i32 [[TMP67]], i32* [[TMP79]], align 4
// CHECK19-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
// CHECK19-NEXT:    store i64 4, i64* [[TMP80]], align 4
// CHECK19-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP81]], align 4
// CHECK19-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP83]], align 4
// CHECK19-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP85]], align 4
// CHECK19-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
// CHECK19-NEXT:    store i64 4, i64* [[TMP86]], align 4
// CHECK19-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP87]], align 4
// CHECK19-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP89]], align 4
// CHECK19-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP91]], align 4
// CHECK19-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
// CHECK19-NEXT:    store i64 [[TMP69]], i64* [[TMP92]], align 4
// CHECK19-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
// CHECK19-NEXT:    store i8* null, i8** [[TMP93]], align 4
// CHECK19-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP97:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK19-NEXT:    [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK19-NEXT:    [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
// CHECK19-NEXT:    [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
// CHECK19-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
// CHECK19-NEXT:    store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK19-NEXT:    [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK19-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
// CHECK19-NEXT:    [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
// CHECK19-NEXT:    [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164.region_id, i32 4, i8** [[TMP94]], i8** [[TMP95]], i64* [[TMP96]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
// CHECK19-NEXT:    br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
// CHECK19:       omp_offload.failed29:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT30]]
// CHECK19:       omp_offload.cont30:
// CHECK19-NEXT:    [[TMP103:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP103]], i32* [[N_CASTED31]], align 4
// CHECK19-NEXT:    [[TMP104:%.*]] = load i32, i32* [[N_CASTED31]], align 4
// CHECK19-NEXT:    [[TMP105:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP106:%.*]] = sext i32 [[TMP105]] to i64
// CHECK19-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP108:%.*]] = bitcast i8** [[TMP107]] to i32*
// CHECK19-NEXT:    store i32 [[TMP104]], i32* [[TMP108]], align 4
// CHECK19-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
// CHECK19-NEXT:    store i32 [[TMP104]], i32* [[TMP110]], align 4
// CHECK19-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
// CHECK19-NEXT:    store i64 4, i64* [[TMP111]], align 4
// CHECK19-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP112]], align 4
// CHECK19-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP114]], align 4
// CHECK19-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP116:%.*]] = bitcast i8** [[TMP115]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP116]], align 4
// CHECK19-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 1
// CHECK19-NEXT:    store i64 4, i64* [[TMP117]], align 4
// CHECK19-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP118]], align 4
// CHECK19-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP120]], align 4
// CHECK19-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP122]], align 4
// CHECK19-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 2
// CHECK19-NEXT:    store i64 [[TMP106]], i64* [[TMP123]], align 4
// CHECK19-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP124]], align 4
// CHECK19-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP128:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP128]], i32* [[DOTCAPTURE_EXPR_37]], align 4
// CHECK19-NEXT:    [[TMP129:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_37]], align 4
// CHECK19-NEXT:    [[SUB39:%.*]] = sub nsw i32 [[TMP129]], 0
// CHECK19-NEXT:    [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
// CHECK19-NEXT:    [[SUB41:%.*]] = sub nsw i32 [[DIV40]], 1
// CHECK19-NEXT:    store i32 [[SUB41]], i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK19-NEXT:    [[TMP130:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK19-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP130]], 1
// CHECK19-NEXT:    [[TMP131:%.*]] = zext i32 [[ADD42]] to i64
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP131]])
// CHECK19-NEXT:    [[TMP132:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169.region_id, i32 3, i8** [[TMP125]], i8** [[TMP126]], i64* [[TMP127]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP133:%.*]] = icmp ne i32 [[TMP132]], 0
// CHECK19-NEXT:    br i1 [[TMP133]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
// CHECK19:       omp_offload.failed43:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i32 [[TMP104]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT44]]
// CHECK19:       omp_offload.cont44:
// CHECK19-NEXT:    [[TMP134:%.*]] = load i32, i32* [[M]], align 4
// CHECK19-NEXT:    store i32 [[TMP134]], i32* [[M_CASTED45]], align 4
// CHECK19-NEXT:    [[TMP135:%.*]] = load i32, i32* [[M_CASTED45]], align 4
// CHECK19-NEXT:    [[TMP136:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP136]], i32* [[N_CASTED46]], align 4
// CHECK19-NEXT:    [[TMP137:%.*]] = load i32, i32* [[N_CASTED46]], align 4
// CHECK19-NEXT:    [[TMP138:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP139:%.*]] = sext i32 [[TMP138]] to i64
// CHECK19-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32*
// CHECK19-NEXT:    store i32 [[TMP135]], i32* [[TMP141]], align 4
// CHECK19-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i32*
// CHECK19-NEXT:    store i32 [[TMP135]], i32* [[TMP143]], align 4
// CHECK19-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
// CHECK19-NEXT:    store i64 4, i64* [[TMP144]], align 4
// CHECK19-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP145]], align 4
// CHECK19-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
// CHECK19-NEXT:    store i32 [[TMP137]], i32* [[TMP147]], align 4
// CHECK19-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
// CHECK19-NEXT:    store i32 [[TMP137]], i32* [[TMP149]], align 4
// CHECK19-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 1
// CHECK19-NEXT:    store i64 4, i64* [[TMP150]], align 4
// CHECK19-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP151]], align 4
// CHECK19-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP153]], align 4
// CHECK19-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP155]], align 4
// CHECK19-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 2
// CHECK19-NEXT:    store i64 4, i64* [[TMP156]], align 4
// CHECK19-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP157]], align 4
// CHECK19-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP159:%.*]] = bitcast i8** [[TMP158]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP159]], align 4
// CHECK19-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP161:%.*]] = bitcast i8** [[TMP160]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP161]], align 4
// CHECK19-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 3
// CHECK19-NEXT:    store i64 [[TMP139]], i64* [[TMP162]], align 4
// CHECK19-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 3
// CHECK19-NEXT:    store i8* null, i8** [[TMP163]], align 4
// CHECK19-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP167:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP167]], i32* [[DOTCAPTURE_EXPR_52]], align 4
// CHECK19-NEXT:    [[TMP168:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_52]], align 4
// CHECK19-NEXT:    [[SUB54:%.*]] = sub nsw i32 [[TMP168]], 0
// CHECK19-NEXT:    [[DIV55:%.*]] = sdiv i32 [[SUB54]], 1
// CHECK19-NEXT:    [[SUB56:%.*]] = sub nsw i32 [[DIV55]], 1
// CHECK19-NEXT:    store i32 [[SUB56]], i32* [[DOTCAPTURE_EXPR_53]], align 4
// CHECK19-NEXT:    [[TMP169:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_53]], align 4
// CHECK19-NEXT:    [[ADD57:%.*]] = add nsw i32 [[TMP169]], 1
// CHECK19-NEXT:    [[TMP170:%.*]] = zext i32 [[ADD57]] to i64
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP170]])
// CHECK19-NEXT:    [[TMP171:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174.region_id, i32 4, i8** [[TMP164]], i8** [[TMP165]], i64* [[TMP166]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP172:%.*]] = icmp ne i32 [[TMP171]], 0
// CHECK19-NEXT:    br i1 [[TMP172]], label [[OMP_OFFLOAD_FAILED58:%.*]], label [[OMP_OFFLOAD_CONT59:%.*]]
// CHECK19:       omp_offload.failed58:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i32 [[TMP135]], i32 [[TMP137]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT59]]
// CHECK19:       omp_offload.cont59:
// CHECK19-NEXT:    [[TMP173:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK19-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP173]])
// CHECK19-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK19-NEXT:    [[TMP174:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP174]])
// CHECK19-NEXT:    [[TMP175:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK19-NEXT:    ret i32 [[TMP175]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK19-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK19-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !14
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK19-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK19-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK19-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK19-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK19-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK19-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK19-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK19-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK19-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !23
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK19-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK19-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK19-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK19-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK19-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK19-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK19-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK19-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-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*, i32, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK19-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK19-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK19-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, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]]), !llvm.access.group !29
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK19-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK19-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
// CHECK19-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK19:       cond.true11:
// CHECK19-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    br label [[COND_END13:%.*]]
// CHECK19:       cond.false12:
// CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    br label [[COND_END13]]
// CHECK19:       cond.end13:
// CHECK19-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
// CHECK19-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
// CHECK19-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK19-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK19-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
// CHECK19-NEXT:    [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD17]], i32* [[I4]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK19-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK19-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK19-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK19-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK19-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK19-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK19-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..8
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK19-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !35
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK19-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK19-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK19-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK19-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK19-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK19-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK19-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK19-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK19-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK19-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK19-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-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*, i32, i32*, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK19-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, i32*, i32, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]]), !llvm.access.group !41
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK19-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK19-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK19-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD9]], i32* [[I4]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK19-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK19-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK19-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK19:       omp.precond.then:
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK19-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK19-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK19-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK19-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK19-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK19-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK19-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK19-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK19-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK19-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[M_CASTED19:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK19-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK19-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK19-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK19:       omp_offload.failed:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK19-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK19-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK19:       omp_offload.failed5:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK19:       omp_offload.cont6:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[M]], align 4
// CHECK19-NEXT:    store i32 [[TMP18]], i32* [[M_CASTED]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
// CHECK19-NEXT:    store i32 [[TMP19]], i32* [[TMP21]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK19-NEXT:    store i32 [[TMP19]], i32* [[TMP23]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP24]], align 4
// CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
// CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
// CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP29]], align 4
// CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK19-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
// CHECK19-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
// CHECK19:       omp_offload.failed11:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
// CHECK19:       omp_offload.cont12:
// CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 4
// CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
// CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP38]], align 4
// CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK19-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK19-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK19:       omp_offload.failed17:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK19:       omp_offload.cont18:
// CHECK19-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK19-NEXT:    store i32 [[TMP43]], i32* [[M_CASTED19]], align 4
// CHECK19-NEXT:    [[TMP44:%.*]] = load i32, i32* [[M_CASTED19]], align 4
// CHECK19-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK19-NEXT:    store i32 [[TMP44]], i32* [[TMP46]], align 4
// CHECK19-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
// CHECK19-NEXT:    store i32 [[TMP44]], i32* [[TMP48]], align 4
// CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP49]], align 4
// CHECK19-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 4
// CHECK19-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 4
// CHECK19-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK19-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
// CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK19-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK19-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
// CHECK19-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK19:       omp_offload.failed24:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i32 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK19:       omp_offload.cont25:
// CHECK19-NEXT:    ret i32 0
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK19-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !47
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK19-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK19-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK19-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !53
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK19-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK19-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK19-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !59
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK19-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK19-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK19-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK19-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK19-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !65
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP66:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK19-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK19-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK19-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK19-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK19-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP69:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK19-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK19-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK19:       cond.true:
// CHECK19-NEXT:    br label [[COND_END:%.*]]
// CHECK19:       cond.false:
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    br label [[COND_END]]
// CHECK19:       cond.end:
// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !71
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP72:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK19-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK19-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK19-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK19-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK19-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP75:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK19-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK19-SAME: () #[[ATTR5:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK19-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@main
// CHECK20-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 4
// CHECK20-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_CASTED3:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
// CHECK20-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_CASTED17:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
// CHECK20-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_CASTED31:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS32:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS33:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS34:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_SIZES35:%.*]] = alloca [3 x i64], align 4
// CHECK20-NEXT:    [[_TMP36:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_37:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[M_CASTED45:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_CASTED46:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS47:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS48:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS49:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_SIZES50:%.*]] = alloca [4 x i64], align 4
// CHECK20-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_52:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_53:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK20-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK20-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK20-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK20-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK20-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK20-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[N_CASTED]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK20-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
// CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP7]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP9]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK20-NEXT:    store i64 4, i64* [[TMP10]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP13]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP15]], align 4
// CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK20-NEXT:    store i64 4, i64* [[TMP16]], align 4
// CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP19]], align 4
// CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP21]], align 4
// CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK20-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP27:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK20-NEXT:    [[TMP30:%.*]] = zext i32 [[ADD]] to i64
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
// CHECK20-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK20-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK20:       omp_offload.failed:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
// CHECK20-NEXT:    [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
// CHECK20-NEXT:    [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK20-NEXT:    [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
// CHECK20-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
// CHECK20-NEXT:    store i32 [[TMP34]], i32* [[TMP38]], align 4
// CHECK20-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
// CHECK20-NEXT:    store i32 [[TMP34]], i32* [[TMP40]], align 4
// CHECK20-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK20-NEXT:    store i64 4, i64* [[TMP41]], align 4
// CHECK20-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP42]], align 4
// CHECK20-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP44]], align 4
// CHECK20-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP46]], align 4
// CHECK20-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
// CHECK20-NEXT:    store i64 4, i64* [[TMP47]], align 4
// CHECK20-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP48]], align 4
// CHECK20-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP50]], align 4
// CHECK20-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP52]], align 4
// CHECK20-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
// CHECK20-NEXT:    store i64 [[TMP36]], i64* [[TMP53]], align 4
// CHECK20-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK20-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP58:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK20-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK20-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
// CHECK20-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
// CHECK20-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
// CHECK20-NEXT:    store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK20-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK20-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
// CHECK20-NEXT:    [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
// CHECK20-NEXT:    [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159.region_id, i32 3, i8** [[TMP55]], i8** [[TMP56]], i64* [[TMP57]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
// CHECK20-NEXT:    br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
// CHECK20:       omp_offload.failed15:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT16]]
// CHECK20:       omp_offload.cont16:
// CHECK20-NEXT:    [[TMP64:%.*]] = load i32, i32* [[M]], align 4
// CHECK20-NEXT:    store i32 [[TMP64]], i32* [[M_CASTED]], align 4
// CHECK20-NEXT:    [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK20-NEXT:    [[TMP66:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
// CHECK20-NEXT:    [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
// CHECK20-NEXT:    [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK20-NEXT:    [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
// CHECK20-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
// CHECK20-NEXT:    store i32 [[TMP65]], i32* [[TMP71]], align 4
// CHECK20-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
// CHECK20-NEXT:    store i32 [[TMP65]], i32* [[TMP73]], align 4
// CHECK20-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
// CHECK20-NEXT:    store i64 4, i64* [[TMP74]], align 4
// CHECK20-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP75]], align 4
// CHECK20-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
// CHECK20-NEXT:    store i32 [[TMP67]], i32* [[TMP77]], align 4
// CHECK20-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
// CHECK20-NEXT:    store i32 [[TMP67]], i32* [[TMP79]], align 4
// CHECK20-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
// CHECK20-NEXT:    store i64 4, i64* [[TMP80]], align 4
// CHECK20-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP81]], align 4
// CHECK20-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP83]], align 4
// CHECK20-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP85]], align 4
// CHECK20-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
// CHECK20-NEXT:    store i64 4, i64* [[TMP86]], align 4
// CHECK20-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP87]], align 4
// CHECK20-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP89]], align 4
// CHECK20-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP91]], align 4
// CHECK20-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
// CHECK20-NEXT:    store i64 [[TMP69]], i64* [[TMP92]], align 4
// CHECK20-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
// CHECK20-NEXT:    store i8* null, i8** [[TMP93]], align 4
// CHECK20-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP97:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK20-NEXT:    [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK20-NEXT:    [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
// CHECK20-NEXT:    [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
// CHECK20-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
// CHECK20-NEXT:    store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK20-NEXT:    [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK20-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
// CHECK20-NEXT:    [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
// CHECK20-NEXT:    [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164.region_id, i32 4, i8** [[TMP94]], i8** [[TMP95]], i64* [[TMP96]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
// CHECK20-NEXT:    br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
// CHECK20:       omp_offload.failed29:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT30]]
// CHECK20:       omp_offload.cont30:
// CHECK20-NEXT:    [[TMP103:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP103]], i32* [[N_CASTED31]], align 4
// CHECK20-NEXT:    [[TMP104:%.*]] = load i32, i32* [[N_CASTED31]], align 4
// CHECK20-NEXT:    [[TMP105:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK20-NEXT:    [[TMP106:%.*]] = sext i32 [[TMP105]] to i64
// CHECK20-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP108:%.*]] = bitcast i8** [[TMP107]] to i32*
// CHECK20-NEXT:    store i32 [[TMP104]], i32* [[TMP108]], align 4
// CHECK20-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
// CHECK20-NEXT:    store i32 [[TMP104]], i32* [[TMP110]], align 4
// CHECK20-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
// CHECK20-NEXT:    store i64 4, i64* [[TMP111]], align 4
// CHECK20-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP112]], align 4
// CHECK20-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP114]], align 4
// CHECK20-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP116:%.*]] = bitcast i8** [[TMP115]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP116]], align 4
// CHECK20-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 1
// CHECK20-NEXT:    store i64 4, i64* [[TMP117]], align 4
// CHECK20-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP118]], align 4
// CHECK20-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP120]], align 4
// CHECK20-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP122]], align 4
// CHECK20-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 2
// CHECK20-NEXT:    store i64 [[TMP106]], i64* [[TMP123]], align 4
// CHECK20-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP124]], align 4
// CHECK20-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP128:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP128]], i32* [[DOTCAPTURE_EXPR_37]], align 4
// CHECK20-NEXT:    [[TMP129:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_37]], align 4
// CHECK20-NEXT:    [[SUB39:%.*]] = sub nsw i32 [[TMP129]], 0
// CHECK20-NEXT:    [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
// CHECK20-NEXT:    [[SUB41:%.*]] = sub nsw i32 [[DIV40]], 1
// CHECK20-NEXT:    store i32 [[SUB41]], i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK20-NEXT:    [[TMP130:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK20-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP130]], 1
// CHECK20-NEXT:    [[TMP131:%.*]] = zext i32 [[ADD42]] to i64
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP131]])
// CHECK20-NEXT:    [[TMP132:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169.region_id, i32 3, i8** [[TMP125]], i8** [[TMP126]], i64* [[TMP127]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP133:%.*]] = icmp ne i32 [[TMP132]], 0
// CHECK20-NEXT:    br i1 [[TMP133]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
// CHECK20:       omp_offload.failed43:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i32 [[TMP104]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT44]]
// CHECK20:       omp_offload.cont44:
// CHECK20-NEXT:    [[TMP134:%.*]] = load i32, i32* [[M]], align 4
// CHECK20-NEXT:    store i32 [[TMP134]], i32* [[M_CASTED45]], align 4
// CHECK20-NEXT:    [[TMP135:%.*]] = load i32, i32* [[M_CASTED45]], align 4
// CHECK20-NEXT:    [[TMP136:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP136]], i32* [[N_CASTED46]], align 4
// CHECK20-NEXT:    [[TMP137:%.*]] = load i32, i32* [[N_CASTED46]], align 4
// CHECK20-NEXT:    [[TMP138:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK20-NEXT:    [[TMP139:%.*]] = sext i32 [[TMP138]] to i64
// CHECK20-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32*
// CHECK20-NEXT:    store i32 [[TMP135]], i32* [[TMP141]], align 4
// CHECK20-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i32*
// CHECK20-NEXT:    store i32 [[TMP135]], i32* [[TMP143]], align 4
// CHECK20-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
// CHECK20-NEXT:    store i64 4, i64* [[TMP144]], align 4
// CHECK20-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP145]], align 4
// CHECK20-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
// CHECK20-NEXT:    store i32 [[TMP137]], i32* [[TMP147]], align 4
// CHECK20-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
// CHECK20-NEXT:    store i32 [[TMP137]], i32* [[TMP149]], align 4
// CHECK20-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 1
// CHECK20-NEXT:    store i64 4, i64* [[TMP150]], align 4
// CHECK20-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP151]], align 4
// CHECK20-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP153]], align 4
// CHECK20-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32*
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP155]], align 4
// CHECK20-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 2
// CHECK20-NEXT:    store i64 4, i64* [[TMP156]], align 4
// CHECK20-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP157]], align 4
// CHECK20-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP159:%.*]] = bitcast i8** [[TMP158]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP159]], align 4
// CHECK20-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP161:%.*]] = bitcast i8** [[TMP160]] to i32**
// CHECK20-NEXT:    store i32* [[VLA]], i32** [[TMP161]], align 4
// CHECK20-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 3
// CHECK20-NEXT:    store i64 [[TMP139]], i64* [[TMP162]], align 4
// CHECK20-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 3
// CHECK20-NEXT:    store i8* null, i8** [[TMP163]], align 4
// CHECK20-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP167:%.*]] = load i32, i32* [[N]], align 4
// CHECK20-NEXT:    store i32 [[TMP167]], i32* [[DOTCAPTURE_EXPR_52]], align 4
// CHECK20-NEXT:    [[TMP168:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_52]], align 4
// CHECK20-NEXT:    [[SUB54:%.*]] = sub nsw i32 [[TMP168]], 0
// CHECK20-NEXT:    [[DIV55:%.*]] = sdiv i32 [[SUB54]], 1
// CHECK20-NEXT:    [[SUB56:%.*]] = sub nsw i32 [[DIV55]], 1
// CHECK20-NEXT:    store i32 [[SUB56]], i32* [[DOTCAPTURE_EXPR_53]], align 4
// CHECK20-NEXT:    [[TMP169:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_53]], align 4
// CHECK20-NEXT:    [[ADD57:%.*]] = add nsw i32 [[TMP169]], 1
// CHECK20-NEXT:    [[TMP170:%.*]] = zext i32 [[ADD57]] to i64
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP170]])
// CHECK20-NEXT:    [[TMP171:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174.region_id, i32 4, i8** [[TMP164]], i8** [[TMP165]], i64* [[TMP166]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP172:%.*]] = icmp ne i32 [[TMP171]], 0
// CHECK20-NEXT:    br i1 [[TMP172]], label [[OMP_OFFLOAD_FAILED58:%.*]], label [[OMP_OFFLOAD_CONT59:%.*]]
// CHECK20:       omp_offload.failed58:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i32 [[TMP135]], i32 [[TMP137]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT59]]
// CHECK20:       omp_offload.cont59:
// CHECK20-NEXT:    [[TMP173:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK20-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP173]])
// CHECK20-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK20-NEXT:    [[TMP174:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP174]])
// CHECK20-NEXT:    [[TMP175:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK20-NEXT:    ret i32 [[TMP175]]
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK20-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK20-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !14
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !14
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !14
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK20-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK20-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK20-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !18
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !18
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK20-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK20-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK20-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK20-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK20-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK20-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK20-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK20-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !23
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK20-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK20-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK20-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !26
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !26
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK20-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK20-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK20-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK20-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK20-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK20-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK20-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-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*, i32, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK20-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK20-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK20-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, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]]), !llvm.access.group !29
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK20-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK20-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK20-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
// CHECK20-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK20:       cond.true11:
// CHECK20-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    br label [[COND_END13:%.*]]
// CHECK20:       cond.false12:
// CHECK20-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    br label [[COND_END13]]
// CHECK20:       cond.end13:
// CHECK20-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
// CHECK20-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
// CHECK20-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK20-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK20-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
// CHECK20-NEXT:    [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD17]], i32* [[I4]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK20-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK20-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !32
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !32
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK20-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK20-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK20-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK20-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK20-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK20-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK20-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK20-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..8
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK20-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !35
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK20-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK20-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK20-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK20:       omp.dispatch.cond:
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK20-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK20-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK20:       omp.dispatch.body:
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK20-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK20-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !38
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !38
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK20-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK20-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK20:       omp.dispatch.inc:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK20:       omp.dispatch.end:
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK20-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK20-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK20-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK20-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK20-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK20-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-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*, i32, i32*, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK20-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK20-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, i32*, i32, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]]), !llvm.access.group !41
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK20-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK20-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK20-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD9]], i32* [[I4]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK20-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK20-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK20-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK20:       omp.precond.then:
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK20-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK20:       omp.dispatch.cond:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK20-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK20-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK20-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK20:       omp.dispatch.body:
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !44
// CHECK20-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK20-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !44
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !44
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !44
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK20-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK20:       omp.dispatch.inc:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK20:       omp.dispatch.end:
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK20-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK20-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK20-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK20-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK20-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK20:       omp.precond.end:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK20-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK20-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[M_CASTED19:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK20-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK20-NEXT:    [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK20-NEXT:    br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK20:       omp_offload.failed:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK20-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK20-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK20:       omp_offload.failed5:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK20:       omp_offload.cont6:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[M]], align 4
// CHECK20-NEXT:    store i32 [[TMP18]], i32* [[M_CASTED]], align 4
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
// CHECK20-NEXT:    store i32 [[TMP19]], i32* [[TMP21]], align 4
// CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK20-NEXT:    store i32 [[TMP19]], i32* [[TMP23]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP24]], align 4
// CHECK20-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
// CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
// CHECK20-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP29]], align 4
// CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK20-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
// CHECK20-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
// CHECK20:       omp_offload.failed11:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
// CHECK20:       omp_offload.cont12:
// CHECK20-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 4
// CHECK20-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
// CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP38]], align 4
// CHECK20-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK20-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK20-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK20:       omp_offload.failed17:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK20:       omp_offload.cont18:
// CHECK20-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK20-NEXT:    store i32 [[TMP43]], i32* [[M_CASTED19]], align 4
// CHECK20-NEXT:    [[TMP44:%.*]] = load i32, i32* [[M_CASTED19]], align 4
// CHECK20-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK20-NEXT:    store i32 [[TMP44]], i32* [[TMP46]], align 4
// CHECK20-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
// CHECK20-NEXT:    store i32 [[TMP44]], i32* [[TMP48]], align 4
// CHECK20-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP49]], align 4
// CHECK20-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 4
// CHECK20-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 4
// CHECK20-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK20-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
// CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
// CHECK20-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK20-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
// CHECK20-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK20:       omp_offload.failed24:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i32 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK20:       omp_offload.cont25:
// CHECK20-NEXT:    ret i32 0
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK20-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !47
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK20-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
// CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !50
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !50
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK20-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK20-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK20-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !53
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !53
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !53
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK20-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !56
// CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !56
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !56
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !56
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK20-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK20-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK20-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !59
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK20-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK20:       omp.dispatch.cond:
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK20:       omp.dispatch.body:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
// CHECK20-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK20-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
// CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !62
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !62
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK20-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK20:       omp.dispatch.inc:
// CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK20-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK20-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK20:       omp.dispatch.end:
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK20-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK20-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !65
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !65
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !65
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP66:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK20-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK20-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK20:       omp.dispatch.cond:
// CHECK20-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK20-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK20-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK20:       omp.dispatch.body:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !68
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !68
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !68
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !68
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK20-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK20-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP69:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK20:       omp.dispatch.inc:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK20:       omp.dispatch.end:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK20-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK20-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK20:       cond.true:
// CHECK20-NEXT:    br label [[COND_END:%.*]]
// CHECK20:       cond.false:
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    br label [[COND_END]]
// CHECK20:       cond.end:
// CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !71
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP72:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK20:       omp.loop.exit:
// CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK20-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK20-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK20:       omp.dispatch.cond:
// CHECK20-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK20-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK20-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK20:       omp.dispatch.body:
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK20-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK20:       omp.inner.for.cond:
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !74
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK20:       omp.inner.for.body:
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !74
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !74
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
// CHECK20-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !74
// CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK20:       omp.body.continue:
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK20:       omp.inner.for.inc:
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK20-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK20-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP75:![0-9]+]]
// CHECK20:       omp.inner.for.end:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK20:       omp.dispatch.inc:
// CHECK20-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK20:       omp.dispatch.end:
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK20-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK20-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK20:       .omp.final.then:
// CHECK20-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK20-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK20:       .omp.final.done:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK20-SAME: () #[[ATTR5:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK20-NEXT:    ret void
//
//
// CHECK21-LABEL: define {{[^@]+}}@main
// CHECK21-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK21-NEXT:  entry:
// CHECK21-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK21-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK21-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK21-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB16:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB17:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I18:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP40:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB46:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB47:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I48:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV51:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I52:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP69:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_70:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_71:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB75:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB76:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I77:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV80:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I81:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_98:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP99:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_100:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_101:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB105:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB106:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I107:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV110:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I111:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK21-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK21-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK21-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK21-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK21-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK21-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK21-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK21-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK21-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK21-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK21-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK21-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK21-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK21-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK21-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK21-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
// CHECK21-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
// CHECK21:       simd.if.then:
// CHECK21-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK21-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK21:       omp.inner.for.cond:
// CHECK21-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK21-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK21:       omp.inner.for.body:
// CHECK21-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK21-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK21-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK21-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK21:       omp.body.continue:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK21:       omp.inner.for.inc:
// CHECK21-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK21-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK21:       omp.inner.for.end:
// CHECK21-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP13]], 0
// CHECK21-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK21-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK21-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK21-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END]]
// CHECK21:       simd.if.end:
// CHECK21-NEXT:    [[TMP14:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK21-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK21-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[TMP15]], 0
// CHECK21-NEXT:    [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
// CHECK21-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
// CHECK21-NEXT:    store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB16]], align 4
// CHECK21-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK21-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_UB17]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I18]], align 4
// CHECK21-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK21-NEXT:    [[CMP19:%.*]] = icmp slt i32 0, [[TMP17]]
// CHECK21-NEXT:    br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END38:%.*]]
// CHECK21:       simd.if.then20:
// CHECK21-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
// CHECK21-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV21]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK21:       omp.inner.for.cond23:
// CHECK21-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK21-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
// CHECK21:       omp.inner.for.body25:
// CHECK21-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK21-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK21-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[IDXPROM28:%.*]] = sext i32 [[TMP22]] to i64
// CHECK21-NEXT:    [[ARRAYIDX29:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM28]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE30:%.*]]
// CHECK21:       omp.body.continue30:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
// CHECK21:       omp.inner.for.inc31:
// CHECK21-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK21-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK21:       omp.inner.for.end33:
// CHECK21-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK21-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK21-NEXT:    [[DIV35:%.*]] = sdiv i32 [[SUB34]], 1
// CHECK21-NEXT:    [[MUL36:%.*]] = mul nsw i32 [[DIV35]], 1
// CHECK21-NEXT:    [[ADD37:%.*]] = add nsw i32 0, [[MUL36]]
// CHECK21-NEXT:    store i32 [[ADD37]], i32* [[I22]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END38]]
// CHECK21:       simd.if.end38:
// CHECK21-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK21-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK21-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK21-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[TMP27]], 0
// CHECK21-NEXT:    [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
// CHECK21-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
// CHECK21-NEXT:    store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB46]], align 4
// CHECK21-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK21-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_UB47]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I48]], align 4
// CHECK21-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK21-NEXT:    [[CMP49:%.*]] = icmp slt i32 0, [[TMP29]]
// CHECK21-NEXT:    br i1 [[CMP49]], label [[SIMD_IF_THEN50:%.*]], label [[SIMD_IF_END68:%.*]]
// CHECK21:       simd.if.then50:
// CHECK21-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_LB46]], align 4
// CHECK21-NEXT:    store i32 [[TMP30]], i32* [[DOTOMP_IV51]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND53:%.*]]
// CHECK21:       omp.inner.for.cond53:
// CHECK21-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_UB47]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[CMP54:%.*]] = icmp sle i32 [[TMP31]], [[TMP32]]
// CHECK21-NEXT:    br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END63:%.*]]
// CHECK21:       omp.inner.for.body55:
// CHECK21-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[MUL56:%.*]] = mul nsw i32 [[TMP33]], 1
// CHECK21-NEXT:    [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
// CHECK21-NEXT:    store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[TMP34:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[IDXPROM58:%.*]] = sext i32 [[TMP34]] to i64
// CHECK21-NEXT:    [[ARRAYIDX59:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM58]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX59]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE60:%.*]]
// CHECK21:       omp.body.continue60:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC61:%.*]]
// CHECK21:       omp.inner.for.inc61:
// CHECK21-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[ADD62:%.*]] = add nsw i32 [[TMP35]], 1
// CHECK21-NEXT:    store i32 [[ADD62]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK21:       omp.inner.for.end63:
// CHECK21-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK21-NEXT:    [[SUB64:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK21-NEXT:    [[DIV65:%.*]] = sdiv i32 [[SUB64]], 1
// CHECK21-NEXT:    [[MUL66:%.*]] = mul nsw i32 [[DIV65]], 1
// CHECK21-NEXT:    [[ADD67:%.*]] = add nsw i32 0, [[MUL66]]
// CHECK21-NEXT:    store i32 [[ADD67]], i32* [[I52]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END68]]
// CHECK21:       simd.if.end68:
// CHECK21-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK21-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK21-NEXT:    [[SUB72:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK21-NEXT:    [[DIV73:%.*]] = sdiv i32 [[SUB72]], 1
// CHECK21-NEXT:    [[SUB74:%.*]] = sub nsw i32 [[DIV73]], 1
// CHECK21-NEXT:    store i32 [[SUB74]], i32* [[DOTCAPTURE_EXPR_71]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB75]], align 4
// CHECK21-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_71]], align 4
// CHECK21-NEXT:    store i32 [[TMP39]], i32* [[DOTOMP_UB76]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I77]], align 4
// CHECK21-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK21-NEXT:    [[CMP78:%.*]] = icmp slt i32 0, [[TMP40]]
// CHECK21-NEXT:    br i1 [[CMP78]], label [[SIMD_IF_THEN79:%.*]], label [[SIMD_IF_END97:%.*]]
// CHECK21:       simd.if.then79:
// CHECK21-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_LB75]], align 4
// CHECK21-NEXT:    store i32 [[TMP41]], i32* [[DOTOMP_IV80]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND82:%.*]]
// CHECK21:       omp.inner.for.cond82:
// CHECK21-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_UB76]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[CMP83:%.*]] = icmp sle i32 [[TMP42]], [[TMP43]]
// CHECK21-NEXT:    br i1 [[CMP83]], label [[OMP_INNER_FOR_BODY84:%.*]], label [[OMP_INNER_FOR_END92:%.*]]
// CHECK21:       omp.inner.for.body84:
// CHECK21-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[MUL85:%.*]] = mul nsw i32 [[TMP44]], 1
// CHECK21-NEXT:    [[ADD86:%.*]] = add nsw i32 0, [[MUL85]]
// CHECK21-NEXT:    store i32 [[ADD86]], i32* [[I81]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[TMP45:%.*]] = load i32, i32* [[I81]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[IDXPROM87:%.*]] = sext i32 [[TMP45]] to i64
// CHECK21-NEXT:    [[ARRAYIDX88:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM87]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX88]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE89:%.*]]
// CHECK21:       omp.body.continue89:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC90:%.*]]
// CHECK21:       omp.inner.for.inc90:
// CHECK21-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[ADD91:%.*]] = add nsw i32 [[TMP46]], 1
// CHECK21-NEXT:    store i32 [[ADD91]], i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND82]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK21:       omp.inner.for.end92:
// CHECK21-NEXT:    [[TMP47:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK21-NEXT:    [[SUB93:%.*]] = sub nsw i32 [[TMP47]], 0
// CHECK21-NEXT:    [[DIV94:%.*]] = sdiv i32 [[SUB93]], 1
// CHECK21-NEXT:    [[MUL95:%.*]] = mul nsw i32 [[DIV94]], 1
// CHECK21-NEXT:    [[ADD96:%.*]] = add nsw i32 0, [[MUL95]]
// CHECK21-NEXT:    store i32 [[ADD96]], i32* [[I81]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END97]]
// CHECK21:       simd.if.end97:
// CHECK21-NEXT:    [[TMP48:%.*]] = load i32, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 [[TMP48]], i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK21-NEXT:    [[TMP49:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP49]], i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK21-NEXT:    [[TMP50:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK21-NEXT:    [[SUB102:%.*]] = sub nsw i32 [[TMP50]], 0
// CHECK21-NEXT:    [[DIV103:%.*]] = sdiv i32 [[SUB102]], 1
// CHECK21-NEXT:    [[SUB104:%.*]] = sub nsw i32 [[DIV103]], 1
// CHECK21-NEXT:    store i32 [[SUB104]], i32* [[DOTCAPTURE_EXPR_101]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB105]], align 4
// CHECK21-NEXT:    [[TMP51:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_101]], align 4
// CHECK21-NEXT:    store i32 [[TMP51]], i32* [[DOTOMP_UB106]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I107]], align 4
// CHECK21-NEXT:    [[TMP52:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK21-NEXT:    [[CMP108:%.*]] = icmp slt i32 0, [[TMP52]]
// CHECK21-NEXT:    br i1 [[CMP108]], label [[SIMD_IF_THEN109:%.*]], label [[SIMD_IF_END127:%.*]]
// CHECK21:       simd.if.then109:
// CHECK21-NEXT:    [[TMP53:%.*]] = load i32, i32* [[DOTOMP_LB105]], align 4
// CHECK21-NEXT:    store i32 [[TMP53]], i32* [[DOTOMP_IV110]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND112:%.*]]
// CHECK21:       omp.inner.for.cond112:
// CHECK21-NEXT:    [[TMP54:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[TMP55:%.*]] = load i32, i32* [[DOTOMP_UB106]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[CMP113:%.*]] = icmp sle i32 [[TMP54]], [[TMP55]]
// CHECK21-NEXT:    br i1 [[CMP113]], label [[OMP_INNER_FOR_BODY114:%.*]], label [[OMP_INNER_FOR_END122:%.*]]
// CHECK21:       omp.inner.for.body114:
// CHECK21-NEXT:    [[TMP56:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[MUL115:%.*]] = mul nsw i32 [[TMP56]], 1
// CHECK21-NEXT:    [[ADD116:%.*]] = add nsw i32 0, [[MUL115]]
// CHECK21-NEXT:    store i32 [[ADD116]], i32* [[I111]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[TMP57:%.*]] = load i32, i32* [[I111]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[IDXPROM117:%.*]] = sext i32 [[TMP57]] to i64
// CHECK21-NEXT:    [[ARRAYIDX118:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM117]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX118]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE119:%.*]]
// CHECK21:       omp.body.continue119:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC120:%.*]]
// CHECK21:       omp.inner.for.inc120:
// CHECK21-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[ADD121:%.*]] = add nsw i32 [[TMP58]], 1
// CHECK21-NEXT:    store i32 [[ADD121]], i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND112]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK21:       omp.inner.for.end122:
// CHECK21-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK21-NEXT:    [[SUB123:%.*]] = sub nsw i32 [[TMP59]], 0
// CHECK21-NEXT:    [[DIV124:%.*]] = sdiv i32 [[SUB123]], 1
// CHECK21-NEXT:    [[MUL125:%.*]] = mul nsw i32 [[DIV124]], 1
// CHECK21-NEXT:    [[ADD126:%.*]] = add nsw i32 0, [[MUL125]]
// CHECK21-NEXT:    store i32 [[ADD126]], i32* [[I111]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END127]]
// CHECK21:       simd.if.end127:
// CHECK21-NEXT:    [[TMP60:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK21-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP60]])
// CHECK21-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK21-NEXT:    [[TMP61:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK21-NEXT:    call void @llvm.stackrestore(i8* [[TMP61]])
// CHECK21-NEXT:    [[TMP62:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK21-NEXT:    ret i32 [[TMP62]]
//
//
// CHECK21-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK21-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
// CHECK21-NEXT:  entry:
// CHECK21-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK21-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB19:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB20:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP34:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB35:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB36:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV37:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I38:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_50:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB52:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB53:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV54:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I55:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK21-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK21-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK21-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK21:       omp.inner.for.cond:
// CHECK21-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK21-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK21:       omp.inner.for.body:
// CHECK21-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK21-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK21-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
// CHECK21-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK21:       omp.body.continue:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK21:       omp.inner.for.inc:
// CHECK21-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK21-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK21:       omp.inner.for.end:
// CHECK21-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB4]], align 4
// CHECK21-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
// CHECK21-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK21:       omp.inner.for.cond7:
// CHECK21-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK21-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
// CHECK21:       omp.inner.for.body9:
// CHECK21-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK21-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK21-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP10]] to i64
// CHECK21-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM12]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX13]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE14:%.*]]
// CHECK21:       omp.body.continue14:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC15:%.*]]
// CHECK21:       omp.inner.for.inc15:
// CHECK21-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK21-NEXT:    store i32 [[ADD16]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK21:       omp.inner.for.end17:
// CHECK21-NEXT:    store i32 10, i32* [[I6]], align 4
// CHECK21-NEXT:    [[TMP12:%.*]] = load i32, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB19]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB20]], align 4
// CHECK21-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB19]], align 4
// CHECK21-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV21]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK21:       omp.inner.for.cond23:
// CHECK21-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB20]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK21-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
// CHECK21:       omp.inner.for.body25:
// CHECK21-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK21-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK21-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[IDXPROM28:%.*]] = sext i32 [[TMP17]] to i64
// CHECK21-NEXT:    [[ARRAYIDX29:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM28]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE30:%.*]]
// CHECK21:       omp.body.continue30:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
// CHECK21:       omp.inner.for.inc31:
// CHECK21-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK21-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK21:       omp.inner.for.end33:
// CHECK21-NEXT:    store i32 10, i32* [[I22]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB35]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB36]], align 4
// CHECK21-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB35]], align 4
// CHECK21-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV37]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND39:%.*]]
// CHECK21:       omp.inner.for.cond39:
// CHECK21-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB36]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[CMP40:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK21-NEXT:    br i1 [[CMP40]], label [[OMP_INNER_FOR_BODY41:%.*]], label [[OMP_INNER_FOR_END49:%.*]]
// CHECK21:       omp.inner.for.body41:
// CHECK21-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[MUL42:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK21-NEXT:    [[ADD43:%.*]] = add nsw i32 0, [[MUL42]]
// CHECK21-NEXT:    store i32 [[ADD43]], i32* [[I38]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I38]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[IDXPROM44:%.*]] = sext i32 [[TMP23]] to i64
// CHECK21-NEXT:    [[ARRAYIDX45:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM44]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX45]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE46:%.*]]
// CHECK21:       omp.body.continue46:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC47:%.*]]
// CHECK21:       omp.inner.for.inc47:
// CHECK21-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[ADD48:%.*]] = add nsw i32 [[TMP24]], 1
// CHECK21-NEXT:    store i32 [[ADD48]], i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND39]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK21:       omp.inner.for.end49:
// CHECK21-NEXT:    store i32 10, i32* [[I38]], align 4
// CHECK21-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_50]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB52]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB53]], align 4
// CHECK21-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB52]], align 4
// CHECK21-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV54]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND56:%.*]]
// CHECK21:       omp.inner.for.cond56:
// CHECK21-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB53]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[CMP57:%.*]] = icmp sle i32 [[TMP27]], [[TMP28]]
// CHECK21-NEXT:    br i1 [[CMP57]], label [[OMP_INNER_FOR_BODY58:%.*]], label [[OMP_INNER_FOR_END66:%.*]]
// CHECK21:       omp.inner.for.body58:
// CHECK21-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[MUL59:%.*]] = mul nsw i32 [[TMP29]], 1
// CHECK21-NEXT:    [[ADD60:%.*]] = add nsw i32 0, [[MUL59]]
// CHECK21-NEXT:    store i32 [[ADD60]], i32* [[I55]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I55]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[IDXPROM61:%.*]] = sext i32 [[TMP30]] to i64
// CHECK21-NEXT:    [[ARRAYIDX62:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM61]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX62]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE63:%.*]]
// CHECK21:       omp.body.continue63:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC64:%.*]]
// CHECK21:       omp.inner.for.inc64:
// CHECK21-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[ADD65:%.*]] = add nsw i32 [[TMP31]], 1
// CHECK21-NEXT:    store i32 [[ADD65]], i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND56]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK21:       omp.inner.for.end66:
// CHECK21-NEXT:    store i32 10, i32* [[I55]], align 4
// CHECK21-NEXT:    ret i32 0
//
//
// CHECK22-LABEL: define {{[^@]+}}@main
// CHECK22-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK22-NEXT:  entry:
// CHECK22-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK22-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK22-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK22-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB16:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB17:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I18:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[_TMP40:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB46:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB47:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I48:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV51:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I52:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[_TMP69:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_70:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_71:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB75:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB76:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I77:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV80:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I81:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_98:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[_TMP99:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_100:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_101:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB105:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB106:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I107:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV110:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I111:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK22-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK22-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK22-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK22-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK22-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK22-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK22-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK22-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK22-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK22-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK22-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK22-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK22-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK22-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK22-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK22-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK22-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK22-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK22-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK22-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK22-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
// CHECK22-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
// CHECK22:       simd.if.then:
// CHECK22-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK22-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK22:       omp.inner.for.cond:
// CHECK22-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK22-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
// CHECK22-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK22-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK22:       omp.inner.for.body:
// CHECK22-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK22-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK22-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK22-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !2
// CHECK22-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !2
// CHECK22-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK22-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK22:       omp.body.continue:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK22:       omp.inner.for.inc:
// CHECK22-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK22-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK22-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK22:       omp.inner.for.end:
// CHECK22-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK22-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP13]], 0
// CHECK22-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK22-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK22-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK22-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK22-NEXT:    br label [[SIMD_IF_END]]
// CHECK22:       simd.if.end:
// CHECK22-NEXT:    [[TMP14:%.*]] = load i32, i32* [[N]], align 4
// CHECK22-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK22-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK22-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[TMP15]], 0
// CHECK22-NEXT:    [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
// CHECK22-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
// CHECK22-NEXT:    store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB16]], align 4
// CHECK22-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK22-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_UB17]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[I18]], align 4
// CHECK22-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK22-NEXT:    [[CMP19:%.*]] = icmp slt i32 0, [[TMP17]]
// CHECK22-NEXT:    br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END38:%.*]]
// CHECK22:       simd.if.then20:
// CHECK22-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
// CHECK22-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV21]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK22:       omp.inner.for.cond23:
// CHECK22-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK22-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !6
// CHECK22-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK22-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
// CHECK22:       omp.inner.for.body25:
// CHECK22-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK22-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK22-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK22-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !6
// CHECK22-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !6
// CHECK22-NEXT:    [[IDXPROM28:%.*]] = sext i32 [[TMP22]] to i64
// CHECK22-NEXT:    [[ARRAYIDX29:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM28]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !6
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE30:%.*]]
// CHECK22:       omp.body.continue30:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
// CHECK22:       omp.inner.for.inc31:
// CHECK22-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK22-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK22-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK22:       omp.inner.for.end33:
// CHECK22-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK22-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK22-NEXT:    [[DIV35:%.*]] = sdiv i32 [[SUB34]], 1
// CHECK22-NEXT:    [[MUL36:%.*]] = mul nsw i32 [[DIV35]], 1
// CHECK22-NEXT:    [[ADD37:%.*]] = add nsw i32 0, [[MUL36]]
// CHECK22-NEXT:    store i32 [[ADD37]], i32* [[I22]], align 4
// CHECK22-NEXT:    br label [[SIMD_IF_END38]]
// CHECK22:       simd.if.end38:
// CHECK22-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK22-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK22-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N]], align 4
// CHECK22-NEXT:    store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK22-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK22-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[TMP27]], 0
// CHECK22-NEXT:    [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
// CHECK22-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
// CHECK22-NEXT:    store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB46]], align 4
// CHECK22-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK22-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_UB47]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[I48]], align 4
// CHECK22-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK22-NEXT:    [[CMP49:%.*]] = icmp slt i32 0, [[TMP29]]
// CHECK22-NEXT:    br i1 [[CMP49]], label [[SIMD_IF_THEN50:%.*]], label [[SIMD_IF_END68:%.*]]
// CHECK22:       simd.if.then50:
// CHECK22-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_LB46]], align 4
// CHECK22-NEXT:    store i32 [[TMP30]], i32* [[DOTOMP_IV51]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND53:%.*]]
// CHECK22:       omp.inner.for.cond53:
// CHECK22-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK22-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_UB47]], align 4, !llvm.access.group !9
// CHECK22-NEXT:    [[CMP54:%.*]] = icmp sle i32 [[TMP31]], [[TMP32]]
// CHECK22-NEXT:    br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END63:%.*]]
// CHECK22:       omp.inner.for.body55:
// CHECK22-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK22-NEXT:    [[MUL56:%.*]] = mul nsw i32 [[TMP33]], 1
// CHECK22-NEXT:    [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
// CHECK22-NEXT:    store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !9
// CHECK22-NEXT:    [[TMP34:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !9
// CHECK22-NEXT:    [[IDXPROM58:%.*]] = sext i32 [[TMP34]] to i64
// CHECK22-NEXT:    [[ARRAYIDX59:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM58]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX59]], align 4, !llvm.access.group !9
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE60:%.*]]
// CHECK22:       omp.body.continue60:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC61:%.*]]
// CHECK22:       omp.inner.for.inc61:
// CHECK22-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK22-NEXT:    [[ADD62:%.*]] = add nsw i32 [[TMP35]], 1
// CHECK22-NEXT:    store i32 [[ADD62]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK22:       omp.inner.for.end63:
// CHECK22-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK22-NEXT:    [[SUB64:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK22-NEXT:    [[DIV65:%.*]] = sdiv i32 [[SUB64]], 1
// CHECK22-NEXT:    [[MUL66:%.*]] = mul nsw i32 [[DIV65]], 1
// CHECK22-NEXT:    [[ADD67:%.*]] = add nsw i32 0, [[MUL66]]
// CHECK22-NEXT:    store i32 [[ADD67]], i32* [[I52]], align 4
// CHECK22-NEXT:    br label [[SIMD_IF_END68]]
// CHECK22:       simd.if.end68:
// CHECK22-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N]], align 4
// CHECK22-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK22-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK22-NEXT:    [[SUB72:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK22-NEXT:    [[DIV73:%.*]] = sdiv i32 [[SUB72]], 1
// CHECK22-NEXT:    [[SUB74:%.*]] = sub nsw i32 [[DIV73]], 1
// CHECK22-NEXT:    store i32 [[SUB74]], i32* [[DOTCAPTURE_EXPR_71]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB75]], align 4
// CHECK22-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_71]], align 4
// CHECK22-NEXT:    store i32 [[TMP39]], i32* [[DOTOMP_UB76]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[I77]], align 4
// CHECK22-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK22-NEXT:    [[CMP78:%.*]] = icmp slt i32 0, [[TMP40]]
// CHECK22-NEXT:    br i1 [[CMP78]], label [[SIMD_IF_THEN79:%.*]], label [[SIMD_IF_END97:%.*]]
// CHECK22:       simd.if.then79:
// CHECK22-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_LB75]], align 4
// CHECK22-NEXT:    store i32 [[TMP41]], i32* [[DOTOMP_IV80]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND82:%.*]]
// CHECK22:       omp.inner.for.cond82:
// CHECK22-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK22-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_UB76]], align 4, !llvm.access.group !12
// CHECK22-NEXT:    [[CMP83:%.*]] = icmp sle i32 [[TMP42]], [[TMP43]]
// CHECK22-NEXT:    br i1 [[CMP83]], label [[OMP_INNER_FOR_BODY84:%.*]], label [[OMP_INNER_FOR_END92:%.*]]
// CHECK22:       omp.inner.for.body84:
// CHECK22-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK22-NEXT:    [[MUL85:%.*]] = mul nsw i32 [[TMP44]], 1
// CHECK22-NEXT:    [[ADD86:%.*]] = add nsw i32 0, [[MUL85]]
// CHECK22-NEXT:    store i32 [[ADD86]], i32* [[I81]], align 4, !llvm.access.group !12
// CHECK22-NEXT:    [[TMP45:%.*]] = load i32, i32* [[I81]], align 4, !llvm.access.group !12
// CHECK22-NEXT:    [[IDXPROM87:%.*]] = sext i32 [[TMP45]] to i64
// CHECK22-NEXT:    [[ARRAYIDX88:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM87]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX88]], align 4, !llvm.access.group !12
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE89:%.*]]
// CHECK22:       omp.body.continue89:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC90:%.*]]
// CHECK22:       omp.inner.for.inc90:
// CHECK22-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK22-NEXT:    [[ADD91:%.*]] = add nsw i32 [[TMP46]], 1
// CHECK22-NEXT:    store i32 [[ADD91]], i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND82]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK22:       omp.inner.for.end92:
// CHECK22-NEXT:    [[TMP47:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK22-NEXT:    [[SUB93:%.*]] = sub nsw i32 [[TMP47]], 0
// CHECK22-NEXT:    [[DIV94:%.*]] = sdiv i32 [[SUB93]], 1
// CHECK22-NEXT:    [[MUL95:%.*]] = mul nsw i32 [[DIV94]], 1
// CHECK22-NEXT:    [[ADD96:%.*]] = add nsw i32 0, [[MUL95]]
// CHECK22-NEXT:    store i32 [[ADD96]], i32* [[I81]], align 4
// CHECK22-NEXT:    br label [[SIMD_IF_END97]]
// CHECK22:       simd.if.end97:
// CHECK22-NEXT:    [[TMP48:%.*]] = load i32, i32* [[M]], align 4
// CHECK22-NEXT:    store i32 [[TMP48]], i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK22-NEXT:    [[TMP49:%.*]] = load i32, i32* [[N]], align 4
// CHECK22-NEXT:    store i32 [[TMP49]], i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK22-NEXT:    [[TMP50:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK22-NEXT:    [[SUB102:%.*]] = sub nsw i32 [[TMP50]], 0
// CHECK22-NEXT:    [[DIV103:%.*]] = sdiv i32 [[SUB102]], 1
// CHECK22-NEXT:    [[SUB104:%.*]] = sub nsw i32 [[DIV103]], 1
// CHECK22-NEXT:    store i32 [[SUB104]], i32* [[DOTCAPTURE_EXPR_101]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB105]], align 4
// CHECK22-NEXT:    [[TMP51:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_101]], align 4
// CHECK22-NEXT:    store i32 [[TMP51]], i32* [[DOTOMP_UB106]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[I107]], align 4
// CHECK22-NEXT:    [[TMP52:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK22-NEXT:    [[CMP108:%.*]] = icmp slt i32 0, [[TMP52]]
// CHECK22-NEXT:    br i1 [[CMP108]], label [[SIMD_IF_THEN109:%.*]], label [[SIMD_IF_END127:%.*]]
// CHECK22:       simd.if.then109:
// CHECK22-NEXT:    [[TMP53:%.*]] = load i32, i32* [[DOTOMP_LB105]], align 4
// CHECK22-NEXT:    store i32 [[TMP53]], i32* [[DOTOMP_IV110]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND112:%.*]]
// CHECK22:       omp.inner.for.cond112:
// CHECK22-NEXT:    [[TMP54:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK22-NEXT:    [[TMP55:%.*]] = load i32, i32* [[DOTOMP_UB106]], align 4, !llvm.access.group !15
// CHECK22-NEXT:    [[CMP113:%.*]] = icmp sle i32 [[TMP54]], [[TMP55]]
// CHECK22-NEXT:    br i1 [[CMP113]], label [[OMP_INNER_FOR_BODY114:%.*]], label [[OMP_INNER_FOR_END122:%.*]]
// CHECK22:       omp.inner.for.body114:
// CHECK22-NEXT:    [[TMP56:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK22-NEXT:    [[MUL115:%.*]] = mul nsw i32 [[TMP56]], 1
// CHECK22-NEXT:    [[ADD116:%.*]] = add nsw i32 0, [[MUL115]]
// CHECK22-NEXT:    store i32 [[ADD116]], i32* [[I111]], align 4, !llvm.access.group !15
// CHECK22-NEXT:    [[TMP57:%.*]] = load i32, i32* [[I111]], align 4, !llvm.access.group !15
// CHECK22-NEXT:    [[IDXPROM117:%.*]] = sext i32 [[TMP57]] to i64
// CHECK22-NEXT:    [[ARRAYIDX118:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM117]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX118]], align 4, !llvm.access.group !15
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE119:%.*]]
// CHECK22:       omp.body.continue119:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC120:%.*]]
// CHECK22:       omp.inner.for.inc120:
// CHECK22-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK22-NEXT:    [[ADD121:%.*]] = add nsw i32 [[TMP58]], 1
// CHECK22-NEXT:    store i32 [[ADD121]], i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND112]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK22:       omp.inner.for.end122:
// CHECK22-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK22-NEXT:    [[SUB123:%.*]] = sub nsw i32 [[TMP59]], 0
// CHECK22-NEXT:    [[DIV124:%.*]] = sdiv i32 [[SUB123]], 1
// CHECK22-NEXT:    [[MUL125:%.*]] = mul nsw i32 [[DIV124]], 1
// CHECK22-NEXT:    [[ADD126:%.*]] = add nsw i32 0, [[MUL125]]
// CHECK22-NEXT:    store i32 [[ADD126]], i32* [[I111]], align 4
// CHECK22-NEXT:    br label [[SIMD_IF_END127]]
// CHECK22:       simd.if.end127:
// CHECK22-NEXT:    [[TMP60:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK22-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP60]])
// CHECK22-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK22-NEXT:    [[TMP61:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK22-NEXT:    call void @llvm.stackrestore(i8* [[TMP61]])
// CHECK22-NEXT:    [[TMP62:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK22-NEXT:    ret i32 [[TMP62]]
//
//
// CHECK22-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK22-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
// CHECK22-NEXT:  entry:
// CHECK22-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK22-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB19:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB20:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[_TMP34:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB35:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB36:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV37:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I38:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTCAPTURE_EXPR_50:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_LB52:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_UB53:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[DOTOMP_IV54:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    [[I55:%.*]] = alloca i32, align 4
// CHECK22-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK22-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK22-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK22-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK22-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK22:       omp.inner.for.cond:
// CHECK22-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK22-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK22-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK22-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK22:       omp.inner.for.body:
// CHECK22-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK22-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK22-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK22-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
// CHECK22-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !18
// CHECK22-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
// CHECK22-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK22:       omp.body.continue:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK22:       omp.inner.for.inc:
// CHECK22-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK22-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK22-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK22:       omp.inner.for.end:
// CHECK22-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
// CHECK22-NEXT:    store i32 9, i32* [[DOTOMP_UB4]], align 4
// CHECK22-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
// CHECK22-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK22:       omp.inner.for.cond7:
// CHECK22-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK22-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !21
// CHECK22-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK22-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
// CHECK22:       omp.inner.for.body9:
// CHECK22-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK22-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK22-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK22-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !21
// CHECK22-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !21
// CHECK22-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP10]] to i64
// CHECK22-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM12]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX13]], align 4, !llvm.access.group !21
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE14:%.*]]
// CHECK22:       omp.body.continue14:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC15:%.*]]
// CHECK22:       omp.inner.for.inc15:
// CHECK22-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK22-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK22-NEXT:    store i32 [[ADD16]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK22:       omp.inner.for.end17:
// CHECK22-NEXT:    store i32 10, i32* [[I6]], align 4
// CHECK22-NEXT:    [[TMP12:%.*]] = load i32, i32* [[M]], align 4
// CHECK22-NEXT:    store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB19]], align 4
// CHECK22-NEXT:    store i32 9, i32* [[DOTOMP_UB20]], align 4
// CHECK22-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB19]], align 4
// CHECK22-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV21]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK22:       omp.inner.for.cond23:
// CHECK22-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK22-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB20]], align 4, !llvm.access.group !24
// CHECK22-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK22-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
// CHECK22:       omp.inner.for.body25:
// CHECK22-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK22-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK22-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK22-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !24
// CHECK22-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !24
// CHECK22-NEXT:    [[IDXPROM28:%.*]] = sext i32 [[TMP17]] to i64
// CHECK22-NEXT:    [[ARRAYIDX29:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM28]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !24
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE30:%.*]]
// CHECK22:       omp.body.continue30:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
// CHECK22:       omp.inner.for.inc31:
// CHECK22-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK22-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK22-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK22:       omp.inner.for.end33:
// CHECK22-NEXT:    store i32 10, i32* [[I22]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB35]], align 4
// CHECK22-NEXT:    store i32 9, i32* [[DOTOMP_UB36]], align 4
// CHECK22-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB35]], align 4
// CHECK22-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV37]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND39:%.*]]
// CHECK22:       omp.inner.for.cond39:
// CHECK22-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK22-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB36]], align 4, !llvm.access.group !27
// CHECK22-NEXT:    [[CMP40:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK22-NEXT:    br i1 [[CMP40]], label [[OMP_INNER_FOR_BODY41:%.*]], label [[OMP_INNER_FOR_END49:%.*]]
// CHECK22:       omp.inner.for.body41:
// CHECK22-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK22-NEXT:    [[MUL42:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK22-NEXT:    [[ADD43:%.*]] = add nsw i32 0, [[MUL42]]
// CHECK22-NEXT:    store i32 [[ADD43]], i32* [[I38]], align 4, !llvm.access.group !27
// CHECK22-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I38]], align 4, !llvm.access.group !27
// CHECK22-NEXT:    [[IDXPROM44:%.*]] = sext i32 [[TMP23]] to i64
// CHECK22-NEXT:    [[ARRAYIDX45:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM44]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX45]], align 4, !llvm.access.group !27
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE46:%.*]]
// CHECK22:       omp.body.continue46:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC47:%.*]]
// CHECK22:       omp.inner.for.inc47:
// CHECK22-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK22-NEXT:    [[ADD48:%.*]] = add nsw i32 [[TMP24]], 1
// CHECK22-NEXT:    store i32 [[ADD48]], i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND39]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK22:       omp.inner.for.end49:
// CHECK22-NEXT:    store i32 10, i32* [[I38]], align 4
// CHECK22-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK22-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_50]], align 4
// CHECK22-NEXT:    store i32 0, i32* [[DOTOMP_LB52]], align 4
// CHECK22-NEXT:    store i32 9, i32* [[DOTOMP_UB53]], align 4
// CHECK22-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB52]], align 4
// CHECK22-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV54]], align 4
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND56:%.*]]
// CHECK22:       omp.inner.for.cond56:
// CHECK22-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK22-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB53]], align 4, !llvm.access.group !30
// CHECK22-NEXT:    [[CMP57:%.*]] = icmp sle i32 [[TMP27]], [[TMP28]]
// CHECK22-NEXT:    br i1 [[CMP57]], label [[OMP_INNER_FOR_BODY58:%.*]], label [[OMP_INNER_FOR_END66:%.*]]
// CHECK22:       omp.inner.for.body58:
// CHECK22-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK22-NEXT:    [[MUL59:%.*]] = mul nsw i32 [[TMP29]], 1
// CHECK22-NEXT:    [[ADD60:%.*]] = add nsw i32 0, [[MUL59]]
// CHECK22-NEXT:    store i32 [[ADD60]], i32* [[I55]], align 4, !llvm.access.group !30
// CHECK22-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I55]], align 4, !llvm.access.group !30
// CHECK22-NEXT:    [[IDXPROM61:%.*]] = sext i32 [[TMP30]] to i64
// CHECK22-NEXT:    [[ARRAYIDX62:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM61]]
// CHECK22-NEXT:    store i32 0, i32* [[ARRAYIDX62]], align 4, !llvm.access.group !30
// CHECK22-NEXT:    br label [[OMP_BODY_CONTINUE63:%.*]]
// CHECK22:       omp.body.continue63:
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_INC64:%.*]]
// CHECK22:       omp.inner.for.inc64:
// CHECK22-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK22-NEXT:    [[ADD65:%.*]] = add nsw i32 [[TMP31]], 1
// CHECK22-NEXT:    store i32 [[ADD65]], i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK22-NEXT:    br label [[OMP_INNER_FOR_COND56]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK22:       omp.inner.for.end66:
// CHECK22-NEXT:    store i32 10, i32* [[I55]], align 4
// CHECK22-NEXT:    ret i32 0
//
//
// CHECK23-LABEL: define {{[^@]+}}@main
// CHECK23-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK23-NEXT:  entry:
// CHECK23-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 4
// CHECK23-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK23-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB16:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB17:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I18:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP39:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB45:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB46:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I47:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV50:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I51:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP67:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_68:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_69:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB73:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB74:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I75:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV78:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I79:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_95:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP96:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_97:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_98:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB102:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB103:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I104:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV107:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I108:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK23-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK23-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK23-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK23-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK23-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK23-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK23-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK23-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
// CHECK23-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK23-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK23-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK23-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK23-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
// CHECK23:       simd.if.then:
// CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK23-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK23:       omp.inner.for.cond:
// CHECK23-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK23-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK23:       omp.inner.for.body:
// CHECK23-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK23-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK23-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP10]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK23:       omp.body.continue:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK23:       omp.inner.for.inc:
// CHECK23-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK23-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK23:       omp.inner.for.end:
// CHECK23-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP12]], 0
// CHECK23-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK23-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK23-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK23-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END]]
// CHECK23:       simd.if.end:
// CHECK23-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP13]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK23-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK23-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[TMP14]], 0
// CHECK23-NEXT:    [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
// CHECK23-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
// CHECK23-NEXT:    store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB16]], align 4
// CHECK23-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK23-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_UB17]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I18]], align 4
// CHECK23-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK23-NEXT:    [[CMP19:%.*]] = icmp slt i32 0, [[TMP16]]
// CHECK23-NEXT:    br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END37:%.*]]
// CHECK23:       simd.if.then20:
// CHECK23-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
// CHECK23-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV21]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK23:       omp.inner.for.cond23:
// CHECK23-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK23-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END32:%.*]]
// CHECK23:       omp.inner.for.body25:
// CHECK23-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK23-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK23-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP21]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX28]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE29:%.*]]
// CHECK23:       omp.body.continue29:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC30:%.*]]
// CHECK23:       omp.inner.for.inc30:
// CHECK23-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK23-NEXT:    store i32 [[ADD31]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK23:       omp.inner.for.end32:
// CHECK23-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK23-NEXT:    [[SUB33:%.*]] = sub nsw i32 [[TMP23]], 0
// CHECK23-NEXT:    [[DIV34:%.*]] = sdiv i32 [[SUB33]], 1
// CHECK23-NEXT:    [[MUL35:%.*]] = mul nsw i32 [[DIV34]], 1
// CHECK23-NEXT:    [[ADD36:%.*]] = add nsw i32 0, [[MUL35]]
// CHECK23-NEXT:    store i32 [[ADD36]], i32* [[I22]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END37]]
// CHECK23:       simd.if.end37:
// CHECK23-NEXT:    [[TMP24:%.*]] = load i32, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 [[TMP24]], i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK23-NEXT:    [[TMP25:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK23-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK23-NEXT:    [[SUB42:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK23-NEXT:    [[DIV43:%.*]] = sdiv i32 [[SUB42]], 1
// CHECK23-NEXT:    [[SUB44:%.*]] = sub nsw i32 [[DIV43]], 1
// CHECK23-NEXT:    store i32 [[SUB44]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB45]], align 4
// CHECK23-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK23-NEXT:    store i32 [[TMP27]], i32* [[DOTOMP_UB46]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I47]], align 4
// CHECK23-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK23-NEXT:    [[CMP48:%.*]] = icmp slt i32 0, [[TMP28]]
// CHECK23-NEXT:    br i1 [[CMP48]], label [[SIMD_IF_THEN49:%.*]], label [[SIMD_IF_END66:%.*]]
// CHECK23:       simd.if.then49:
// CHECK23-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_LB45]], align 4
// CHECK23-NEXT:    store i32 [[TMP29]], i32* [[DOTOMP_IV50]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND52:%.*]]
// CHECK23:       omp.inner.for.cond52:
// CHECK23-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_UB46]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[CMP53:%.*]] = icmp sle i32 [[TMP30]], [[TMP31]]
// CHECK23-NEXT:    br i1 [[CMP53]], label [[OMP_INNER_FOR_BODY54:%.*]], label [[OMP_INNER_FOR_END61:%.*]]
// CHECK23:       omp.inner.for.body54:
// CHECK23-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[MUL55:%.*]] = mul nsw i32 [[TMP32]], 1
// CHECK23-NEXT:    [[ADD56:%.*]] = add nsw i32 0, [[MUL55]]
// CHECK23-NEXT:    store i32 [[ADD56]], i32* [[I51]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[TMP33:%.*]] = load i32, i32* [[I51]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[ARRAYIDX57:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP33]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX57]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE58:%.*]]
// CHECK23:       omp.body.continue58:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC59:%.*]]
// CHECK23:       omp.inner.for.inc59:
// CHECK23-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP34]], 1
// CHECK23-NEXT:    store i32 [[ADD60]], i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND52]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK23:       omp.inner.for.end61:
// CHECK23-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK23-NEXT:    [[SUB62:%.*]] = sub nsw i32 [[TMP35]], 0
// CHECK23-NEXT:    [[DIV63:%.*]] = sdiv i32 [[SUB62]], 1
// CHECK23-NEXT:    [[MUL64:%.*]] = mul nsw i32 [[DIV63]], 1
// CHECK23-NEXT:    [[ADD65:%.*]] = add nsw i32 0, [[MUL64]]
// CHECK23-NEXT:    store i32 [[ADD65]], i32* [[I51]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END66]]
// CHECK23:       simd.if.end66:
// CHECK23-NEXT:    [[TMP36:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK23-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK23-NEXT:    [[SUB70:%.*]] = sub nsw i32 [[TMP37]], 0
// CHECK23-NEXT:    [[DIV71:%.*]] = sdiv i32 [[SUB70]], 1
// CHECK23-NEXT:    [[SUB72:%.*]] = sub nsw i32 [[DIV71]], 1
// CHECK23-NEXT:    store i32 [[SUB72]], i32* [[DOTCAPTURE_EXPR_69]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB73]], align 4
// CHECK23-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_69]], align 4
// CHECK23-NEXT:    store i32 [[TMP38]], i32* [[DOTOMP_UB74]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I75]], align 4
// CHECK23-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK23-NEXT:    [[CMP76:%.*]] = icmp slt i32 0, [[TMP39]]
// CHECK23-NEXT:    br i1 [[CMP76]], label [[SIMD_IF_THEN77:%.*]], label [[SIMD_IF_END94:%.*]]
// CHECK23:       simd.if.then77:
// CHECK23-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_LB73]], align 4
// CHECK23-NEXT:    store i32 [[TMP40]], i32* [[DOTOMP_IV78]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND80:%.*]]
// CHECK23:       omp.inner.for.cond80:
// CHECK23-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_UB74]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[CMP81:%.*]] = icmp sle i32 [[TMP41]], [[TMP42]]
// CHECK23-NEXT:    br i1 [[CMP81]], label [[OMP_INNER_FOR_BODY82:%.*]], label [[OMP_INNER_FOR_END89:%.*]]
// CHECK23:       omp.inner.for.body82:
// CHECK23-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[MUL83:%.*]] = mul nsw i32 [[TMP43]], 1
// CHECK23-NEXT:    [[ADD84:%.*]] = add nsw i32 0, [[MUL83]]
// CHECK23-NEXT:    store i32 [[ADD84]], i32* [[I79]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[TMP44:%.*]] = load i32, i32* [[I79]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[ARRAYIDX85:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP44]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX85]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE86:%.*]]
// CHECK23:       omp.body.continue86:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC87:%.*]]
// CHECK23:       omp.inner.for.inc87:
// CHECK23-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[ADD88:%.*]] = add nsw i32 [[TMP45]], 1
// CHECK23-NEXT:    store i32 [[ADD88]], i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND80]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK23:       omp.inner.for.end89:
// CHECK23-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK23-NEXT:    [[SUB90:%.*]] = sub nsw i32 [[TMP46]], 0
// CHECK23-NEXT:    [[DIV91:%.*]] = sdiv i32 [[SUB90]], 1
// CHECK23-NEXT:    [[MUL92:%.*]] = mul nsw i32 [[DIV91]], 1
// CHECK23-NEXT:    [[ADD93:%.*]] = add nsw i32 0, [[MUL92]]
// CHECK23-NEXT:    store i32 [[ADD93]], i32* [[I79]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END94]]
// CHECK23:       simd.if.end94:
// CHECK23-NEXT:    [[TMP47:%.*]] = load i32, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 [[TMP47]], i32* [[DOTCAPTURE_EXPR_95]], align 4
// CHECK23-NEXT:    [[TMP48:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP48]], i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK23-NEXT:    [[TMP49:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK23-NEXT:    [[SUB99:%.*]] = sub nsw i32 [[TMP49]], 0
// CHECK23-NEXT:    [[DIV100:%.*]] = sdiv i32 [[SUB99]], 1
// CHECK23-NEXT:    [[SUB101:%.*]] = sub nsw i32 [[DIV100]], 1
// CHECK23-NEXT:    store i32 [[SUB101]], i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB102]], align 4
// CHECK23-NEXT:    [[TMP50:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK23-NEXT:    store i32 [[TMP50]], i32* [[DOTOMP_UB103]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I104]], align 4
// CHECK23-NEXT:    [[TMP51:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK23-NEXT:    [[CMP105:%.*]] = icmp slt i32 0, [[TMP51]]
// CHECK23-NEXT:    br i1 [[CMP105]], label [[SIMD_IF_THEN106:%.*]], label [[SIMD_IF_END123:%.*]]
// CHECK23:       simd.if.then106:
// CHECK23-NEXT:    [[TMP52:%.*]] = load i32, i32* [[DOTOMP_LB102]], align 4
// CHECK23-NEXT:    store i32 [[TMP52]], i32* [[DOTOMP_IV107]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND109:%.*]]
// CHECK23:       omp.inner.for.cond109:
// CHECK23-NEXT:    [[TMP53:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[TMP54:%.*]] = load i32, i32* [[DOTOMP_UB103]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[CMP110:%.*]] = icmp sle i32 [[TMP53]], [[TMP54]]
// CHECK23-NEXT:    br i1 [[CMP110]], label [[OMP_INNER_FOR_BODY111:%.*]], label [[OMP_INNER_FOR_END118:%.*]]
// CHECK23:       omp.inner.for.body111:
// CHECK23-NEXT:    [[TMP55:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[MUL112:%.*]] = mul nsw i32 [[TMP55]], 1
// CHECK23-NEXT:    [[ADD113:%.*]] = add nsw i32 0, [[MUL112]]
// CHECK23-NEXT:    store i32 [[ADD113]], i32* [[I108]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[TMP56:%.*]] = load i32, i32* [[I108]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[ARRAYIDX114:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP56]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX114]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE115:%.*]]
// CHECK23:       omp.body.continue115:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC116:%.*]]
// CHECK23:       omp.inner.for.inc116:
// CHECK23-NEXT:    [[TMP57:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[ADD117:%.*]] = add nsw i32 [[TMP57]], 1
// CHECK23-NEXT:    store i32 [[ADD117]], i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND109]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK23:       omp.inner.for.end118:
// CHECK23-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK23-NEXT:    [[SUB119:%.*]] = sub nsw i32 [[TMP58]], 0
// CHECK23-NEXT:    [[DIV120:%.*]] = sdiv i32 [[SUB119]], 1
// CHECK23-NEXT:    [[MUL121:%.*]] = mul nsw i32 [[DIV120]], 1
// CHECK23-NEXT:    [[ADD122:%.*]] = add nsw i32 0, [[MUL121]]
// CHECK23-NEXT:    store i32 [[ADD122]], i32* [[I108]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END123]]
// CHECK23:       simd.if.end123:
// CHECK23-NEXT:    [[TMP59:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK23-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP59]])
// CHECK23-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK23-NEXT:    [[TMP60:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK23-NEXT:    call void @llvm.stackrestore(i8* [[TMP60]])
// CHECK23-NEXT:    [[TMP61:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK23-NEXT:    ret i32 [[TMP61]]
//
//
// CHECK23-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK23-SAME: (i32 [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
// CHECK23-NEXT:  entry:
// CHECK23-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK23-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP17:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB18:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB19:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV20:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I21:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB33:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB34:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV35:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I36:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_47:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP48:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB49:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB50:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV51:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I52:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK23-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK23-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK23-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK23:       omp.inner.for.cond:
// CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK23-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK23:       omp.inner.for.body:
// CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK23-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK23-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP4]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK23:       omp.body.continue:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK23:       omp.inner.for.inc:
// CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK23-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
// CHECK23:       omp.inner.for.end:
// CHECK23-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB4]], align 4
// CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
// CHECK23-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK23:       omp.inner.for.cond7:
// CHECK23-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK23-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
// CHECK23:       omp.inner.for.body9:
// CHECK23-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK23-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK23-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP10]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX12]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE13:%.*]]
// CHECK23:       omp.body.continue13:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC14:%.*]]
// CHECK23:       omp.inner.for.inc14:
// CHECK23-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK23-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK23:       omp.inner.for.end16:
// CHECK23-NEXT:    store i32 10, i32* [[I6]], align 4
// CHECK23-NEXT:    [[TMP12:%.*]] = load i32, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB18]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB19]], align 4
// CHECK23-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB18]], align 4
// CHECK23-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV20]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
// CHECK23:       omp.inner.for.cond22:
// CHECK23-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB19]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK23-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END31:%.*]]
// CHECK23:       omp.inner.for.body24:
// CHECK23-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK23-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
// CHECK23-NEXT:    store i32 [[ADD26]], i32* [[I21]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I21]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[ARRAYIDX27:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP17]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX27]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE28:%.*]]
// CHECK23:       omp.body.continue28:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC29:%.*]]
// CHECK23:       omp.inner.for.inc29:
// CHECK23-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[ADD30:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK23-NEXT:    store i32 [[ADD30]], i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK23:       omp.inner.for.end31:
// CHECK23-NEXT:    store i32 10, i32* [[I21]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB33]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB34]], align 4
// CHECK23-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB33]], align 4
// CHECK23-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV35]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND37:%.*]]
// CHECK23:       omp.inner.for.cond37:
// CHECK23-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB34]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK23-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY39:%.*]], label [[OMP_INNER_FOR_END46:%.*]]
// CHECK23:       omp.inner.for.body39:
// CHECK23-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[MUL40:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK23-NEXT:    [[ADD41:%.*]] = add nsw i32 0, [[MUL40]]
// CHECK23-NEXT:    store i32 [[ADD41]], i32* [[I36]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I36]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[ARRAYIDX42:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP23]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX42]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE43:%.*]]
// CHECK23:       omp.body.continue43:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC44:%.*]]
// CHECK23:       omp.inner.for.inc44:
// CHECK23-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[ADD45:%.*]] = add nsw i32 [[TMP24]], 1
// CHECK23-NEXT:    store i32 [[ADD45]], i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND37]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK23:       omp.inner.for.end46:
// CHECK23-NEXT:    store i32 10, i32* [[I36]], align 4
// CHECK23-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_47]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB49]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB50]], align 4
// CHECK23-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB49]], align 4
// CHECK23-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV51]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND53:%.*]]
// CHECK23:       omp.inner.for.cond53:
// CHECK23-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB50]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[CMP54:%.*]] = icmp sle i32 [[TMP27]], [[TMP28]]
// CHECK23-NEXT:    br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END62:%.*]]
// CHECK23:       omp.inner.for.body55:
// CHECK23-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[MUL56:%.*]] = mul nsw i32 [[TMP29]], 1
// CHECK23-NEXT:    [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
// CHECK23-NEXT:    store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[ARRAYIDX58:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP30]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX58]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE59:%.*]]
// CHECK23:       omp.body.continue59:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC60:%.*]]
// CHECK23:       omp.inner.for.inc60:
// CHECK23-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[ADD61:%.*]] = add nsw i32 [[TMP31]], 1
// CHECK23-NEXT:    store i32 [[ADD61]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK23:       omp.inner.for.end62:
// CHECK23-NEXT:    store i32 10, i32* [[I52]], align 4
// CHECK23-NEXT:    ret i32 0
//
//
// CHECK24-LABEL: define {{[^@]+}}@main
// CHECK24-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK24-NEXT:  entry:
// CHECK24-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 4
// CHECK24-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK24-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB16:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB17:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I18:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[_TMP39:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB45:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB46:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I47:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV50:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I51:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[_TMP67:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_68:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_69:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB73:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB74:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I75:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV78:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I79:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_95:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[_TMP96:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_97:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_98:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB102:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB103:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I104:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV107:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I108:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK24-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK24-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK24-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK24-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK24-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK24-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK24-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK24-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK24-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK24-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK24-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK24-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK24-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
// CHECK24-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK24-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK24-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK24-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK24-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK24-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK24-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK24-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
// CHECK24:       simd.if.then:
// CHECK24-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK24-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK24:       omp.inner.for.cond:
// CHECK24-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK24-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
// CHECK24-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK24-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK24:       omp.inner.for.body:
// CHECK24-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK24-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK24-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK24-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !3
// CHECK24-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !3
// CHECK24-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP10]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK24:       omp.body.continue:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK24:       omp.inner.for.inc:
// CHECK24-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK24-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK24-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK24:       omp.inner.for.end:
// CHECK24-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK24-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP12]], 0
// CHECK24-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK24-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK24-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK24-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK24-NEXT:    br label [[SIMD_IF_END]]
// CHECK24:       simd.if.end:
// CHECK24-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N]], align 4
// CHECK24-NEXT:    store i32 [[TMP13]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK24-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK24-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[TMP14]], 0
// CHECK24-NEXT:    [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
// CHECK24-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
// CHECK24-NEXT:    store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB16]], align 4
// CHECK24-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK24-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_UB17]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[I18]], align 4
// CHECK24-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK24-NEXT:    [[CMP19:%.*]] = icmp slt i32 0, [[TMP16]]
// CHECK24-NEXT:    br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END37:%.*]]
// CHECK24:       simd.if.then20:
// CHECK24-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
// CHECK24-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV21]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK24:       omp.inner.for.cond23:
// CHECK24-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK24-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !7
// CHECK24-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK24-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END32:%.*]]
// CHECK24:       omp.inner.for.body25:
// CHECK24-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK24-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK24-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK24-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !7
// CHECK24-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !7
// CHECK24-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP21]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX28]], align 4, !llvm.access.group !7
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE29:%.*]]
// CHECK24:       omp.body.continue29:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC30:%.*]]
// CHECK24:       omp.inner.for.inc30:
// CHECK24-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK24-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK24-NEXT:    store i32 [[ADD31]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK24:       omp.inner.for.end32:
// CHECK24-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK24-NEXT:    [[SUB33:%.*]] = sub nsw i32 [[TMP23]], 0
// CHECK24-NEXT:    [[DIV34:%.*]] = sdiv i32 [[SUB33]], 1
// CHECK24-NEXT:    [[MUL35:%.*]] = mul nsw i32 [[DIV34]], 1
// CHECK24-NEXT:    [[ADD36:%.*]] = add nsw i32 0, [[MUL35]]
// CHECK24-NEXT:    store i32 [[ADD36]], i32* [[I22]], align 4
// CHECK24-NEXT:    br label [[SIMD_IF_END37]]
// CHECK24:       simd.if.end37:
// CHECK24-NEXT:    [[TMP24:%.*]] = load i32, i32* [[M]], align 4
// CHECK24-NEXT:    store i32 [[TMP24]], i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK24-NEXT:    [[TMP25:%.*]] = load i32, i32* [[N]], align 4
// CHECK24-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK24-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK24-NEXT:    [[SUB42:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK24-NEXT:    [[DIV43:%.*]] = sdiv i32 [[SUB42]], 1
// CHECK24-NEXT:    [[SUB44:%.*]] = sub nsw i32 [[DIV43]], 1
// CHECK24-NEXT:    store i32 [[SUB44]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB45]], align 4
// CHECK24-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK24-NEXT:    store i32 [[TMP27]], i32* [[DOTOMP_UB46]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[I47]], align 4
// CHECK24-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK24-NEXT:    [[CMP48:%.*]] = icmp slt i32 0, [[TMP28]]
// CHECK24-NEXT:    br i1 [[CMP48]], label [[SIMD_IF_THEN49:%.*]], label [[SIMD_IF_END66:%.*]]
// CHECK24:       simd.if.then49:
// CHECK24-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_LB45]], align 4
// CHECK24-NEXT:    store i32 [[TMP29]], i32* [[DOTOMP_IV50]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND52:%.*]]
// CHECK24:       omp.inner.for.cond52:
// CHECK24-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK24-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_UB46]], align 4, !llvm.access.group !10
// CHECK24-NEXT:    [[CMP53:%.*]] = icmp sle i32 [[TMP30]], [[TMP31]]
// CHECK24-NEXT:    br i1 [[CMP53]], label [[OMP_INNER_FOR_BODY54:%.*]], label [[OMP_INNER_FOR_END61:%.*]]
// CHECK24:       omp.inner.for.body54:
// CHECK24-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK24-NEXT:    [[MUL55:%.*]] = mul nsw i32 [[TMP32]], 1
// CHECK24-NEXT:    [[ADD56:%.*]] = add nsw i32 0, [[MUL55]]
// CHECK24-NEXT:    store i32 [[ADD56]], i32* [[I51]], align 4, !llvm.access.group !10
// CHECK24-NEXT:    [[TMP33:%.*]] = load i32, i32* [[I51]], align 4, !llvm.access.group !10
// CHECK24-NEXT:    [[ARRAYIDX57:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP33]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX57]], align 4, !llvm.access.group !10
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE58:%.*]]
// CHECK24:       omp.body.continue58:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC59:%.*]]
// CHECK24:       omp.inner.for.inc59:
// CHECK24-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK24-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP34]], 1
// CHECK24-NEXT:    store i32 [[ADD60]], i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND52]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK24:       omp.inner.for.end61:
// CHECK24-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK24-NEXT:    [[SUB62:%.*]] = sub nsw i32 [[TMP35]], 0
// CHECK24-NEXT:    [[DIV63:%.*]] = sdiv i32 [[SUB62]], 1
// CHECK24-NEXT:    [[MUL64:%.*]] = mul nsw i32 [[DIV63]], 1
// CHECK24-NEXT:    [[ADD65:%.*]] = add nsw i32 0, [[MUL64]]
// CHECK24-NEXT:    store i32 [[ADD65]], i32* [[I51]], align 4
// CHECK24-NEXT:    br label [[SIMD_IF_END66]]
// CHECK24:       simd.if.end66:
// CHECK24-NEXT:    [[TMP36:%.*]] = load i32, i32* [[N]], align 4
// CHECK24-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK24-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK24-NEXT:    [[SUB70:%.*]] = sub nsw i32 [[TMP37]], 0
// CHECK24-NEXT:    [[DIV71:%.*]] = sdiv i32 [[SUB70]], 1
// CHECK24-NEXT:    [[SUB72:%.*]] = sub nsw i32 [[DIV71]], 1
// CHECK24-NEXT:    store i32 [[SUB72]], i32* [[DOTCAPTURE_EXPR_69]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB73]], align 4
// CHECK24-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_69]], align 4
// CHECK24-NEXT:    store i32 [[TMP38]], i32* [[DOTOMP_UB74]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[I75]], align 4
// CHECK24-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK24-NEXT:    [[CMP76:%.*]] = icmp slt i32 0, [[TMP39]]
// CHECK24-NEXT:    br i1 [[CMP76]], label [[SIMD_IF_THEN77:%.*]], label [[SIMD_IF_END94:%.*]]
// CHECK24:       simd.if.then77:
// CHECK24-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_LB73]], align 4
// CHECK24-NEXT:    store i32 [[TMP40]], i32* [[DOTOMP_IV78]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND80:%.*]]
// CHECK24:       omp.inner.for.cond80:
// CHECK24-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK24-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_UB74]], align 4, !llvm.access.group !13
// CHECK24-NEXT:    [[CMP81:%.*]] = icmp sle i32 [[TMP41]], [[TMP42]]
// CHECK24-NEXT:    br i1 [[CMP81]], label [[OMP_INNER_FOR_BODY82:%.*]], label [[OMP_INNER_FOR_END89:%.*]]
// CHECK24:       omp.inner.for.body82:
// CHECK24-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK24-NEXT:    [[MUL83:%.*]] = mul nsw i32 [[TMP43]], 1
// CHECK24-NEXT:    [[ADD84:%.*]] = add nsw i32 0, [[MUL83]]
// CHECK24-NEXT:    store i32 [[ADD84]], i32* [[I79]], align 4, !llvm.access.group !13
// CHECK24-NEXT:    [[TMP44:%.*]] = load i32, i32* [[I79]], align 4, !llvm.access.group !13
// CHECK24-NEXT:    [[ARRAYIDX85:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP44]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX85]], align 4, !llvm.access.group !13
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE86:%.*]]
// CHECK24:       omp.body.continue86:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC87:%.*]]
// CHECK24:       omp.inner.for.inc87:
// CHECK24-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK24-NEXT:    [[ADD88:%.*]] = add nsw i32 [[TMP45]], 1
// CHECK24-NEXT:    store i32 [[ADD88]], i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND80]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK24:       omp.inner.for.end89:
// CHECK24-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK24-NEXT:    [[SUB90:%.*]] = sub nsw i32 [[TMP46]], 0
// CHECK24-NEXT:    [[DIV91:%.*]] = sdiv i32 [[SUB90]], 1
// CHECK24-NEXT:    [[MUL92:%.*]] = mul nsw i32 [[DIV91]], 1
// CHECK24-NEXT:    [[ADD93:%.*]] = add nsw i32 0, [[MUL92]]
// CHECK24-NEXT:    store i32 [[ADD93]], i32* [[I79]], align 4
// CHECK24-NEXT:    br label [[SIMD_IF_END94]]
// CHECK24:       simd.if.end94:
// CHECK24-NEXT:    [[TMP47:%.*]] = load i32, i32* [[M]], align 4
// CHECK24-NEXT:    store i32 [[TMP47]], i32* [[DOTCAPTURE_EXPR_95]], align 4
// CHECK24-NEXT:    [[TMP48:%.*]] = load i32, i32* [[N]], align 4
// CHECK24-NEXT:    store i32 [[TMP48]], i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK24-NEXT:    [[TMP49:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK24-NEXT:    [[SUB99:%.*]] = sub nsw i32 [[TMP49]], 0
// CHECK24-NEXT:    [[DIV100:%.*]] = sdiv i32 [[SUB99]], 1
// CHECK24-NEXT:    [[SUB101:%.*]] = sub nsw i32 [[DIV100]], 1
// CHECK24-NEXT:    store i32 [[SUB101]], i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB102]], align 4
// CHECK24-NEXT:    [[TMP50:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK24-NEXT:    store i32 [[TMP50]], i32* [[DOTOMP_UB103]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[I104]], align 4
// CHECK24-NEXT:    [[TMP51:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK24-NEXT:    [[CMP105:%.*]] = icmp slt i32 0, [[TMP51]]
// CHECK24-NEXT:    br i1 [[CMP105]], label [[SIMD_IF_THEN106:%.*]], label [[SIMD_IF_END123:%.*]]
// CHECK24:       simd.if.then106:
// CHECK24-NEXT:    [[TMP52:%.*]] = load i32, i32* [[DOTOMP_LB102]], align 4
// CHECK24-NEXT:    store i32 [[TMP52]], i32* [[DOTOMP_IV107]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND109:%.*]]
// CHECK24:       omp.inner.for.cond109:
// CHECK24-NEXT:    [[TMP53:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK24-NEXT:    [[TMP54:%.*]] = load i32, i32* [[DOTOMP_UB103]], align 4, !llvm.access.group !16
// CHECK24-NEXT:    [[CMP110:%.*]] = icmp sle i32 [[TMP53]], [[TMP54]]
// CHECK24-NEXT:    br i1 [[CMP110]], label [[OMP_INNER_FOR_BODY111:%.*]], label [[OMP_INNER_FOR_END118:%.*]]
// CHECK24:       omp.inner.for.body111:
// CHECK24-NEXT:    [[TMP55:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK24-NEXT:    [[MUL112:%.*]] = mul nsw i32 [[TMP55]], 1
// CHECK24-NEXT:    [[ADD113:%.*]] = add nsw i32 0, [[MUL112]]
// CHECK24-NEXT:    store i32 [[ADD113]], i32* [[I108]], align 4, !llvm.access.group !16
// CHECK24-NEXT:    [[TMP56:%.*]] = load i32, i32* [[I108]], align 4, !llvm.access.group !16
// CHECK24-NEXT:    [[ARRAYIDX114:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP56]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX114]], align 4, !llvm.access.group !16
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE115:%.*]]
// CHECK24:       omp.body.continue115:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC116:%.*]]
// CHECK24:       omp.inner.for.inc116:
// CHECK24-NEXT:    [[TMP57:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK24-NEXT:    [[ADD117:%.*]] = add nsw i32 [[TMP57]], 1
// CHECK24-NEXT:    store i32 [[ADD117]], i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND109]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK24:       omp.inner.for.end118:
// CHECK24-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK24-NEXT:    [[SUB119:%.*]] = sub nsw i32 [[TMP58]], 0
// CHECK24-NEXT:    [[DIV120:%.*]] = sdiv i32 [[SUB119]], 1
// CHECK24-NEXT:    [[MUL121:%.*]] = mul nsw i32 [[DIV120]], 1
// CHECK24-NEXT:    [[ADD122:%.*]] = add nsw i32 0, [[MUL121]]
// CHECK24-NEXT:    store i32 [[ADD122]], i32* [[I108]], align 4
// CHECK24-NEXT:    br label [[SIMD_IF_END123]]
// CHECK24:       simd.if.end123:
// CHECK24-NEXT:    [[TMP59:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK24-NEXT:    [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP59]])
// CHECK24-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK24-NEXT:    [[TMP60:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK24-NEXT:    call void @llvm.stackrestore(i8* [[TMP60]])
// CHECK24-NEXT:    [[TMP61:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK24-NEXT:    ret i32 [[TMP61]]
//
//
// CHECK24-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK24-SAME: (i32 [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
// CHECK24-NEXT:  entry:
// CHECK24-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK24-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[_TMP17:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB18:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB19:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV20:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I21:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB33:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB34:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV35:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I36:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTCAPTURE_EXPR_47:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[_TMP48:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_LB49:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_UB50:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[DOTOMP_IV51:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    [[I52:%.*]] = alloca i32, align 4
// CHECK24-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK24-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK24-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK24-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK24-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK24:       omp.inner.for.cond:
// CHECK24-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK24-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
// CHECK24-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK24-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK24:       omp.inner.for.body:
// CHECK24-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK24-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK24-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK24-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
// CHECK24-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !19
// CHECK24-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP4]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK24:       omp.body.continue:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK24:       omp.inner.for.inc:
// CHECK24-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK24-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK24-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
// CHECK24:       omp.inner.for.end:
// CHECK24-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
// CHECK24-NEXT:    store i32 9, i32* [[DOTOMP_UB4]], align 4
// CHECK24-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
// CHECK24-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK24:       omp.inner.for.cond7:
// CHECK24-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK24-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !22
// CHECK24-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK24-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
// CHECK24:       omp.inner.for.body9:
// CHECK24-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK24-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK24-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK24-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !22
// CHECK24-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !22
// CHECK24-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP10]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX12]], align 4, !llvm.access.group !22
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE13:%.*]]
// CHECK24:       omp.body.continue13:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC14:%.*]]
// CHECK24:       omp.inner.for.inc14:
// CHECK24-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK24-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK24-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK24:       omp.inner.for.end16:
// CHECK24-NEXT:    store i32 10, i32* [[I6]], align 4
// CHECK24-NEXT:    [[TMP12:%.*]] = load i32, i32* [[M]], align 4
// CHECK24-NEXT:    store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB18]], align 4
// CHECK24-NEXT:    store i32 9, i32* [[DOTOMP_UB19]], align 4
// CHECK24-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB18]], align 4
// CHECK24-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV20]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
// CHECK24:       omp.inner.for.cond22:
// CHECK24-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK24-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB19]], align 4, !llvm.access.group !25
// CHECK24-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK24-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END31:%.*]]
// CHECK24:       omp.inner.for.body24:
// CHECK24-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK24-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK24-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
// CHECK24-NEXT:    store i32 [[ADD26]], i32* [[I21]], align 4, !llvm.access.group !25
// CHECK24-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I21]], align 4, !llvm.access.group !25
// CHECK24-NEXT:    [[ARRAYIDX27:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP17]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX27]], align 4, !llvm.access.group !25
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE28:%.*]]
// CHECK24:       omp.body.continue28:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC29:%.*]]
// CHECK24:       omp.inner.for.inc29:
// CHECK24-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK24-NEXT:    [[ADD30:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK24-NEXT:    store i32 [[ADD30]], i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK24:       omp.inner.for.end31:
// CHECK24-NEXT:    store i32 10, i32* [[I21]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB33]], align 4
// CHECK24-NEXT:    store i32 9, i32* [[DOTOMP_UB34]], align 4
// CHECK24-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB33]], align 4
// CHECK24-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV35]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND37:%.*]]
// CHECK24:       omp.inner.for.cond37:
// CHECK24-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK24-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB34]], align 4, !llvm.access.group !28
// CHECK24-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK24-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY39:%.*]], label [[OMP_INNER_FOR_END46:%.*]]
// CHECK24:       omp.inner.for.body39:
// CHECK24-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK24-NEXT:    [[MUL40:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK24-NEXT:    [[ADD41:%.*]] = add nsw i32 0, [[MUL40]]
// CHECK24-NEXT:    store i32 [[ADD41]], i32* [[I36]], align 4, !llvm.access.group !28
// CHECK24-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I36]], align 4, !llvm.access.group !28
// CHECK24-NEXT:    [[ARRAYIDX42:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP23]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX42]], align 4, !llvm.access.group !28
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE43:%.*]]
// CHECK24:       omp.body.continue43:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC44:%.*]]
// CHECK24:       omp.inner.for.inc44:
// CHECK24-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK24-NEXT:    [[ADD45:%.*]] = add nsw i32 [[TMP24]], 1
// CHECK24-NEXT:    store i32 [[ADD45]], i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND37]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK24:       omp.inner.for.end46:
// CHECK24-NEXT:    store i32 10, i32* [[I36]], align 4
// CHECK24-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK24-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_47]], align 4
// CHECK24-NEXT:    store i32 0, i32* [[DOTOMP_LB49]], align 4
// CHECK24-NEXT:    store i32 9, i32* [[DOTOMP_UB50]], align 4
// CHECK24-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB49]], align 4
// CHECK24-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV51]], align 4
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND53:%.*]]
// CHECK24:       omp.inner.for.cond53:
// CHECK24-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK24-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB50]], align 4, !llvm.access.group !31
// CHECK24-NEXT:    [[CMP54:%.*]] = icmp sle i32 [[TMP27]], [[TMP28]]
// CHECK24-NEXT:    br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END62:%.*]]
// CHECK24:       omp.inner.for.body55:
// CHECK24-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK24-NEXT:    [[MUL56:%.*]] = mul nsw i32 [[TMP29]], 1
// CHECK24-NEXT:    [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
// CHECK24-NEXT:    store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !31
// CHECK24-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !31
// CHECK24-NEXT:    [[ARRAYIDX58:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP30]]
// CHECK24-NEXT:    store i32 0, i32* [[ARRAYIDX58]], align 4, !llvm.access.group !31
// CHECK24-NEXT:    br label [[OMP_BODY_CONTINUE59:%.*]]
// CHECK24:       omp.body.continue59:
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_INC60:%.*]]
// CHECK24:       omp.inner.for.inc60:
// CHECK24-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK24-NEXT:    [[ADD61:%.*]] = add nsw i32 [[TMP31]], 1
// CHECK24-NEXT:    store i32 [[ADD61]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK24-NEXT:    br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK24:       omp.inner.for.end62:
// CHECK24-NEXT:    store i32 10, i32* [[I52]], align 4
// CHECK24-NEXT:    ret i32 0
//