nuudlman/llvm: clang/test/OpenMP/teams_distribute_parallel_for_schedule

// 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 -no-opaque-pointers -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK1
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK3

// RUN: %clang_cc1 -no-opaque-pointers -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=CHECK5
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK5
// RUN: %clang_cc1 -no-opaque-pointers -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=CHECK7
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 --implicit-check-not="{{__kmpc|__tgt}}"
#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
    for(int i = 0; i < X; i++) {
      a[i] = (T)0;
    }
    #pragma omp target
    #pragma omp teams distribute parallel for schedule(static)
    for(int i = 0; i < X; i++) {
      a[i] = (T)0;
    }
    #pragma omp target
    #pragma omp teams distribute parallel for schedule(static, X/2)
    for(int i = 0; i < X; i++) {
      a[i] = (T)0;
    }

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

    #pragma omp target
    #pragma omp teams distribute parallel for 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 -no-opaque-pointers -DCK2 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK13
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK13
// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK15
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK15

// RUN: %clang_cc1 -no-opaque-pointers -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=CHECK17
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK17
// RUN: %clang_cc1 -no-opaque-pointers -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=CHECK19
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK19

// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 --implicit-check-not="{{__kmpc|__tgt}}"
#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
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
#pragma omp target
#pragma omp teams distribute parallel for schedule(static)
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
#pragma omp target
#pragma omp teams distribute parallel for schedule(static, m)
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
#pragma omp target
#pragma omp teams distribute parallel for schedule(dynamic)
  for(int i = 0; i < n; i++) {
    a[i] = (T)0;
  }
#pragma omp target
#pragma omp teams distribute parallel for 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
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
#pragma omp target
#pragma omp teams distribute parallel for dist_schedule(static)
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
#pragma omp target
#pragma omp teams distribute parallel for dist_schedule(static, m)
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
#pragma omp target
#pragma omp teams distribute parallel for schedule(dynamic)
  for(int i = 0; i < n; i++) {
    a[i] = 0;
  }
#pragma omp target
#pragma omp teams distribute parallel for 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 noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
// CHECK1-NEXT:    ret i32 [[CALL]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK1-SAME: (%struct.SS* noundef 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_BASEPTRS11:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[_TMP30:%.*]] = 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK1-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP13]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP14]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK1-NEXT:    store i64 123, i64* [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK1-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK1-NEXT:    br i1 [[TMP17]], 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_l35(%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:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], 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_PTRS4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], 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_BASEPTRS3]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
// CHECK1-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
// CHECK1-NEXT:    store i64 123, i64* [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
// CHECK1-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK1-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK1:       omp_offload.failed8:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK1:       omp_offload.cont9:
// CHECK1-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], 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_PTRS12]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 8
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], 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_BASEPTRS11]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP43]], align 4
// CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK1-NEXT:    store i32 1, i32* [[TMP44]], align 4
// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP41]], i8*** [[TMP45]], align 8
// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP42]], i8*** [[TMP46]], align 8
// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 8
// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 8
// CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP49]], align 8
// CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP50]], align 8
// CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK1-NEXT:    store i64 123, i64* [[TMP51]], align 8
// CHECK1-NEXT:    [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK1-NEXT:    [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
// CHECK1-NEXT:    br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK1:       omp_offload.failed16:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK1:       omp_offload.cont17:
// CHECK1-NEXT:    [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to %struct.SS**
// CHECK1-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP55]], align 8
// CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A18]], [123 x i32]** [[TMP57]], align 8
// CHECK1-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP58]], align 8
// CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS23:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP61]], align 4
// CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 1
// CHECK1-NEXT:    store i32 1, i32* [[TMP62]], align 4
// CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP59]], i8*** [[TMP63]], align 8
// CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP60]], i8*** [[TMP64]], align 8
// CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP65]], align 8
// CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP66]], align 8
// CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP67]], align 8
// CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP68]], align 8
// CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 8
// CHECK1-NEXT:    store i64 123, i64* [[TMP69]], align 8
// CHECK1-NEXT:    [[TMP70:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]])
// CHECK1-NEXT:    [[TMP71:%.*]] = icmp ne i32 [[TMP70]], 0
// CHECK1-NEXT:    br i1 [[TMP71]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK1:       omp_offload.failed24:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK1:       omp_offload.cont25:
// CHECK1-NEXT:    [[A26:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to %struct.SS**
// CHECK1-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP73]], align 8
// CHECK1-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [123 x i32]**
// CHECK1-NEXT:    store [123 x i32]* [[A26]], [123 x i32]** [[TMP75]], align 8
// CHECK1-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP76]], align 8
// CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS31:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP79]], align 4
// CHECK1-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 1
// CHECK1-NEXT:    store i32 1, i32* [[TMP80]], align 4
// CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP77]], i8*** [[TMP81]], align 8
// CHECK1-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP78]], i8*** [[TMP82]], align 8
// CHECK1-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP83]], align 8
// CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP84]], align 8
// CHECK1-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP85]], align 8
// CHECK1-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP86]], align 8
// CHECK1-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 8
// CHECK1-NEXT:    store i64 123, i64* [[TMP87]], align 8
// CHECK1-NEXT:    [[TMP88:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]])
// CHECK1-NEXT:    [[TMP89:%.*]] = icmp ne i32 [[TMP88]], 0
// CHECK1-NEXT:    br i1 [[TMP89]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK1:       omp_offload.failed32:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK1:       omp_offload.cont33:
// CHECK1-NEXT:    [[A34:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A34]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP90:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    ret i32 [[TMP90]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
// CHECK1-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// 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
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
// CHECK1-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// 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
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
// CHECK1-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
// 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
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_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:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
// CHECK1-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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 1073741859, 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 !10
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
// 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 !10
// 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 !10
// 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 !10
// 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 !10
// 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 !10
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![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:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
// CHECK1-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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 1073741859, 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 !13
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
// 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 !13
// 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 !13
// 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 !13
// 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 !13
// 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 !13
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![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:    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
//
//
// 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 noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
// CHECK3-NEXT:    ret i32 [[CALL]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK3-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[_TMP30:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 4
// 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 4
// CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP4]], align 4
// 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK3-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP13]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP14]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK3-NEXT:    store i64 123, i64* [[TMP15]], align 8
// CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK3-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK3-NEXT:    br i1 [[TMP17]], 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_l35(%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:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 4
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
// CHECK3-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 4
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 4
// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 4
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP31]], align 4
// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP32]], align 4
// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
// CHECK3-NEXT:    store i64 123, i64* [[TMP33]], align 8
// CHECK3-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
// CHECK3-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK3-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK3:       omp_offload.failed8:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK3:       omp_offload.cont9:
// CHECK3-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 4
// CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP40]], align 4
// CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP43]], align 4
// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK3-NEXT:    store i32 1, i32* [[TMP44]], align 4
// CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP41]], i8*** [[TMP45]], align 4
// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP42]], i8*** [[TMP46]], align 4
// CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 4
// CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 4
// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP49]], align 4
// CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP50]], align 4
// CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK3-NEXT:    store i64 123, i64* [[TMP51]], align 8
// CHECK3-NEXT:    [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK3-NEXT:    [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
// CHECK3-NEXT:    br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK3:       omp_offload.failed16:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK3:       omp_offload.cont17:
// CHECK3-NEXT:    [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP55]], align 4
// CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A18]], [123 x i32]** [[TMP57]], align 4
// CHECK3-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP58]], align 4
// CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS23:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP61]], align 4
// CHECK3-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 1
// CHECK3-NEXT:    store i32 1, i32* [[TMP62]], align 4
// CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP59]], i8*** [[TMP63]], align 4
// CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP60]], i8*** [[TMP64]], align 4
// CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP65]], align 4
// CHECK3-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP66]], align 4
// CHECK3-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP67]], align 4
// CHECK3-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP68]], align 4
// CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 8
// CHECK3-NEXT:    store i64 123, i64* [[TMP69]], align 8
// CHECK3-NEXT:    [[TMP70:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]])
// CHECK3-NEXT:    [[TMP71:%.*]] = icmp ne i32 [[TMP70]], 0
// CHECK3-NEXT:    br i1 [[TMP71]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK3:       omp_offload.failed24:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK3:       omp_offload.cont25:
// CHECK3-NEXT:    [[A26:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP73]], align 4
// CHECK3-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [123 x i32]**
// CHECK3-NEXT:    store [123 x i32]* [[A26]], [123 x i32]** [[TMP75]], align 4
// CHECK3-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP76]], align 4
// CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS31:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK3-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP79]], align 4
// CHECK3-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 1
// CHECK3-NEXT:    store i32 1, i32* [[TMP80]], align 4
// CHECK3-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP77]], i8*** [[TMP81]], align 4
// CHECK3-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP78]], i8*** [[TMP82]], align 4
// CHECK3-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP83]], align 4
// CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP84]], align 4
// CHECK3-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP85]], align 4
// CHECK3-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP86]], align 4
// CHECK3-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 8
// CHECK3-NEXT:    store i64 123, i64* [[TMP87]], align 8
// CHECK3-NEXT:    [[TMP88:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]])
// CHECK3-NEXT:    [[TMP89:%.*]] = icmp ne i32 [[TMP88]], 0
// CHECK3-NEXT:    br i1 [[TMP89]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK3:       omp_offload.failed32:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK3:       omp_offload.cont33:
// CHECK3-NEXT:    [[A34:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A34]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP90:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    ret i32 [[TMP90]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
// CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 4
// 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
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// 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
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK3-NEXT:    br i1 [[CMP1]], 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
// 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
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
// CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 4
// 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
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// 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
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK3-NEXT:    br i1 [[CMP1]], 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
// 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
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
// CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 4
// 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
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// 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 i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// 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 [ [[TMP7]], [[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:    [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK3-NEXT:    br i1 [[CMP1]], 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
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    br i1 [[CMP2]], 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
// 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
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK3-NEXT:    store i32 [[ADD4]], 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:    [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK3-NEXT:    store i32 [[ADD5]], 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:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
// CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 4
// 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
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load 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 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK3-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, 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 !11
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
// 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 !11
// 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 !11
// 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 !11
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11
// 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 !11
// CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK3-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![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:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
// CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// 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 4
// 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
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// 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 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load 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 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK3-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, 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 !14
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
// 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 !14
// 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 !14
// 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 !14
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14
// 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 !14
// CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK3-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![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:    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
//
//
// 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 noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
// CHECK5-NEXT:    ret i32 [[CALL]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK5-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 8
// CHECK5-NEXT:    [[_TMP30:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 8
// 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 8
// CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP4]], align 8
// 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK5-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK5-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK5-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK5-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 8
// CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK5-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 8
// CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8
// CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8
// CHECK5-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK5-NEXT:    store i8** null, i8*** [[TMP13]], align 8
// CHECK5-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK5-NEXT:    store i8** null, i8*** [[TMP14]], align 8
// CHECK5-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK5-NEXT:    store i64 123, i64* [[TMP15]], align 8
// CHECK5-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK5-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK5-NEXT:    br i1 [[TMP17]], 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_l35(%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:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
// CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 8
// CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK5-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK5-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
// CHECK5-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
// CHECK5-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK5-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
// CHECK5-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 8
// CHECK5-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
// CHECK5-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 8
// CHECK5-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 8
// CHECK5-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 8
// CHECK5-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
// CHECK5-NEXT:    store i8** null, i8*** [[TMP31]], align 8
// CHECK5-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
// CHECK5-NEXT:    store i8** null, i8*** [[TMP32]], align 8
// CHECK5-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
// CHECK5-NEXT:    store i64 123, i64* [[TMP33]], align 8
// CHECK5-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
// CHECK5-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK5-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK5:       omp_offload.failed8:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK5:       omp_offload.cont9:
// CHECK5-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
// CHECK5-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 8
// CHECK5-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP40]], align 8
// CHECK5-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK5-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK5-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK5-NEXT:    store i32 1, i32* [[TMP43]], align 4
// CHECK5-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK5-NEXT:    store i32 1, i32* [[TMP44]], align 4
// CHECK5-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK5-NEXT:    store i8** [[TMP41]], i8*** [[TMP45]], align 8
// CHECK5-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK5-NEXT:    store i8** [[TMP42]], i8*** [[TMP46]], align 8
// CHECK5-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 8
// CHECK5-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 8
// CHECK5-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK5-NEXT:    store i8** null, i8*** [[TMP49]], align 8
// CHECK5-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK5-NEXT:    store i8** null, i8*** [[TMP50]], align 8
// CHECK5-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK5-NEXT:    store i64 123, i64* [[TMP51]], align 8
// CHECK5-NEXT:    [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK5-NEXT:    [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
// CHECK5-NEXT:    br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK5:       omp_offload.failed16:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK5:       omp_offload.cont17:
// CHECK5-NEXT:    [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP55]], align 8
// CHECK5-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A18]], [123 x i32]** [[TMP57]], align 8
// CHECK5-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i64 0, i64 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP58]], align 8
// CHECK5-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK5-NEXT:    [[KERNEL_ARGS23:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK5-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 0
// CHECK5-NEXT:    store i32 1, i32* [[TMP61]], align 4
// CHECK5-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 1
// CHECK5-NEXT:    store i32 1, i32* [[TMP62]], align 4
// CHECK5-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 2
// CHECK5-NEXT:    store i8** [[TMP59]], i8*** [[TMP63]], align 8
// CHECK5-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 3
// CHECK5-NEXT:    store i8** [[TMP60]], i8*** [[TMP64]], align 8
// CHECK5-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 4
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP65]], align 8
// CHECK5-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 5
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP66]], align 8
// CHECK5-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 6
// CHECK5-NEXT:    store i8** null, i8*** [[TMP67]], align 8
// CHECK5-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 7
// CHECK5-NEXT:    store i8** null, i8*** [[TMP68]], align 8
// CHECK5-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 8
// CHECK5-NEXT:    store i64 123, i64* [[TMP69]], align 8
// CHECK5-NEXT:    [[TMP70:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]])
// CHECK5-NEXT:    [[TMP71:%.*]] = icmp ne i32 [[TMP70]], 0
// CHECK5-NEXT:    br i1 [[TMP71]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK5:       omp_offload.failed24:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK5:       omp_offload.cont25:
// CHECK5-NEXT:    [[A26:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP73]], align 8
// CHECK5-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A26]], [123 x i32]** [[TMP75]], align 8
// CHECK5-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i64 0, i64 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP76]], align 8
// CHECK5-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK5-NEXT:    [[KERNEL_ARGS31:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK5-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 0
// CHECK5-NEXT:    store i32 1, i32* [[TMP79]], align 4
// CHECK5-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 1
// CHECK5-NEXT:    store i32 1, i32* [[TMP80]], align 4
// CHECK5-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 2
// CHECK5-NEXT:    store i8** [[TMP77]], i8*** [[TMP81]], align 8
// CHECK5-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 3
// CHECK5-NEXT:    store i8** [[TMP78]], i8*** [[TMP82]], align 8
// CHECK5-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 4
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP83]], align 8
// CHECK5-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 5
// CHECK5-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP84]], align 8
// CHECK5-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 6
// CHECK5-NEXT:    store i8** null, i8*** [[TMP85]], align 8
// CHECK5-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 7
// CHECK5-NEXT:    store i8** null, i8*** [[TMP86]], align 8
// CHECK5-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 8
// CHECK5-NEXT:    store i64 123, i64* [[TMP87]], align 8
// CHECK5-NEXT:    [[TMP88:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]])
// CHECK5-NEXT:    [[TMP89:%.*]] = icmp ne i32 [[TMP88]], 0
// CHECK5-NEXT:    br i1 [[TMP89]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK5:       omp_offload.failed32:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK5:       omp_offload.cont33:
// CHECK5-NEXT:    [[A34:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A34]], i64 0, i64 0
// CHECK5-NEXT:    [[TMP90:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK5-NEXT:    ret i32 [[TMP90]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 8
// 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
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK5-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// 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*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], 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 8
// 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
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK5-NEXT:    br i1 [[CMP2]], 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
// 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
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 8
// 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
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK5-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// 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*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], 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 8
// 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
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK5-NEXT:    br i1 [[CMP2]], 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
// 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
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 8
// 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
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK5-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// 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*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], 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 8
// 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 i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// 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 [ [[CONV3]], [[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:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    br i1 [[CMP4]], 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
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    br i1 [[CMP5]], 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
// 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
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
// CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK5-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK5-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK5-NEXT:    store i32 [[ADD7]], 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:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK5-NEXT:    store i32 [[ADD8]], 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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 8
// 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
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK5-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// 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*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], 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 8
// 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 !10
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
// 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 !10
// 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 !10
// 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 !10
// CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !10
// 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 !10
// CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK5-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// 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 8
// 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
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK5-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// 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*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// 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 8
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], 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 8
// 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 !13
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
// 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 !13
// 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 !13
// 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 !13
// CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// 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:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 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_DISPATCH_INC:%.*]]
// CHECK5:       omp.dispatch.inc:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK5:       omp.dispatch.end:
// 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
//
//
// 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 noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
// CHECK7-NEXT:    ret i32 [[CALL]]
//
//
// CHECK7-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK7-SAME: (%struct.SS* noundef 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_BASEPTRS11:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 4
// CHECK7-NEXT:    [[_TMP30:%.*]] = 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK7-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK7-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK7-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK7-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK7-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK7-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 4
// CHECK7-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK7-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
// CHECK7-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
// CHECK7-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK7-NEXT:    store i8** null, i8*** [[TMP13]], align 4
// CHECK7-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK7-NEXT:    store i8** null, i8*** [[TMP14]], align 4
// CHECK7-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK7-NEXT:    store i64 123, i64* [[TMP15]], align 8
// CHECK7-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK7-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK7-NEXT:    br i1 [[TMP17]], 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_l35(%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:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], 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_PTRS4]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 4
// CHECK7-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], 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_BASEPTRS3]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK7-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK7-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
// CHECK7-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK7-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
// CHECK7-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK7-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
// CHECK7-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 4
// CHECK7-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
// CHECK7-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 4
// CHECK7-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 4
// CHECK7-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 4
// CHECK7-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
// CHECK7-NEXT:    store i8** null, i8*** [[TMP31]], align 4
// CHECK7-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
// CHECK7-NEXT:    store i8** null, i8*** [[TMP32]], align 4
// CHECK7-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
// CHECK7-NEXT:    store i64 123, i64* [[TMP33]], align 8
// CHECK7-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
// CHECK7-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK7-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK7:       omp_offload.failed8:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK7:       omp_offload.cont9:
// CHECK7-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], 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_PTRS12]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 4
// CHECK7-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], 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_BASEPTRS11]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK7-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK7-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK7-NEXT:    store i32 1, i32* [[TMP43]], align 4
// CHECK7-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK7-NEXT:    store i32 1, i32* [[TMP44]], align 4
// CHECK7-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK7-NEXT:    store i8** [[TMP41]], i8*** [[TMP45]], align 4
// CHECK7-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK7-NEXT:    store i8** [[TMP42]], i8*** [[TMP46]], align 4
// CHECK7-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 4
// CHECK7-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 4
// CHECK7-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK7-NEXT:    store i8** null, i8*** [[TMP49]], align 4
// CHECK7-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK7-NEXT:    store i8** null, i8*** [[TMP50]], align 4
// CHECK7-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK7-NEXT:    store i64 123, i64* [[TMP51]], align 8
// CHECK7-NEXT:    [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK7-NEXT:    [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
// CHECK7-NEXT:    br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK7:       omp_offload.failed16:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK7:       omp_offload.cont17:
// CHECK7-NEXT:    [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to %struct.SS**
// CHECK7-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP55]], align 4
// CHECK7-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A18]], [123 x i32]** [[TMP57]], align 4
// CHECK7-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
// CHECK7-NEXT:    store i8* null, i8** [[TMP58]], align 4
// CHECK7-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK7-NEXT:    [[KERNEL_ARGS23:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK7-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 0
// CHECK7-NEXT:    store i32 1, i32* [[TMP61]], align 4
// CHECK7-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 1
// CHECK7-NEXT:    store i32 1, i32* [[TMP62]], align 4
// CHECK7-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 2
// CHECK7-NEXT:    store i8** [[TMP59]], i8*** [[TMP63]], align 4
// CHECK7-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 3
// CHECK7-NEXT:    store i8** [[TMP60]], i8*** [[TMP64]], align 4
// CHECK7-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 4
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP65]], align 4
// CHECK7-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 5
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP66]], align 4
// CHECK7-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 6
// CHECK7-NEXT:    store i8** null, i8*** [[TMP67]], align 4
// CHECK7-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 7
// CHECK7-NEXT:    store i8** null, i8*** [[TMP68]], align 4
// CHECK7-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 8
// CHECK7-NEXT:    store i64 123, i64* [[TMP69]], align 8
// CHECK7-NEXT:    [[TMP70:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]])
// CHECK7-NEXT:    [[TMP71:%.*]] = icmp ne i32 [[TMP70]], 0
// CHECK7-NEXT:    br i1 [[TMP71]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK7:       omp_offload.failed24:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK7:       omp_offload.cont25:
// CHECK7-NEXT:    [[A26:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to %struct.SS**
// CHECK7-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP73]], align 4
// CHECK7-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [123 x i32]**
// CHECK7-NEXT:    store [123 x i32]* [[A26]], [123 x i32]** [[TMP75]], align 4
// CHECK7-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i32 0, i32 0
// CHECK7-NEXT:    store i8* null, i8** [[TMP76]], align 4
// CHECK7-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK7-NEXT:    [[KERNEL_ARGS31:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK7-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 0
// CHECK7-NEXT:    store i32 1, i32* [[TMP79]], align 4
// CHECK7-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 1
// CHECK7-NEXT:    store i32 1, i32* [[TMP80]], align 4
// CHECK7-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 2
// CHECK7-NEXT:    store i8** [[TMP77]], i8*** [[TMP81]], align 4
// CHECK7-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 3
// CHECK7-NEXT:    store i8** [[TMP78]], i8*** [[TMP82]], align 4
// CHECK7-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 4
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP83]], align 4
// CHECK7-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 5
// CHECK7-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP84]], align 4
// CHECK7-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 6
// CHECK7-NEXT:    store i8** null, i8*** [[TMP85]], align 4
// CHECK7-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 7
// CHECK7-NEXT:    store i8** null, i8*** [[TMP86]], align 4
// CHECK7-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 8
// CHECK7-NEXT:    store i64 123, i64* [[TMP87]], align 8
// CHECK7-NEXT:    [[TMP88:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]])
// CHECK7-NEXT:    [[TMP89:%.*]] = icmp ne i32 [[TMP88]], 0
// CHECK7-NEXT:    br i1 [[TMP89]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK7:       omp_offload.failed32:
// CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK7:       omp_offload.cont33:
// CHECK7-NEXT:    [[A34:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A34]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP90:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK7-NEXT:    ret i32 [[TMP90]]
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
// CHECK7-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK7-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
// CHECK7-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK7-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
// CHECK7-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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
// CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK7-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK7-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
// CHECK7-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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 !11
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
// 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 !11
// 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 !11
// 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 !11
// 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 !11
// 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 !11
// CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
// CHECK7-SAME: (%struct.SS* noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[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
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK7:       omp.inner.for.inc:
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK7-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[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 !14
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
// 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 !14
// 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 !14
// 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 !14
// 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 !14
// 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 !14
// CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![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:    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
//
//
// CHECK13-LABEL: define {{[^@]+}}@main
// CHECK13-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[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_CASTED20:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS23:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_SIZES25:%.*]] = alloca [4 x i64], align 8
// CHECK13-NEXT:    [[_TMP26:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_28:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_CASTED36:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS38:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS39:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS40:%.*]] = alloca [3 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_SIZES41:%.*]] = alloca [3 x i64], align 8
// CHECK13-NEXT:    [[_TMP42:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_43:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_44:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[M_CASTED52:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[N_CASTED54:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS56:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS57:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS58:%.*]] = alloca [4 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_SIZES59:%.*]] = alloca [4 x i64], align 8
// CHECK13-NEXT:    [[_TMP60:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_61:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_62:%.*]] = 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:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK13-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
// CHECK13-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK13-NEXT:    store i64 [[TMP4]], i64* [[TMP8]], align 8
// CHECK13-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK13-NEXT:    store i64 [[TMP4]], 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 i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK13-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP18]], align 8
// CHECK13-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP20]], align 8
// CHECK13-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK13-NEXT:    store i64 [[TMP5]], i64* [[TMP21]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP27]], 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:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK13-NEXT:    [[TMP29:%.*]] = zext i32 [[ADD]] to i64
// CHECK13-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK13-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP30]], align 4
// CHECK13-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK13-NEXT:    store i32 3, i32* [[TMP31]], align 4
// CHECK13-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP23]], i8*** [[TMP32]], align 8
// CHECK13-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP24]], i8*** [[TMP33]], align 8
// CHECK13-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK13-NEXT:    store i64* [[TMP25]], i64** [[TMP34]], align 8
// CHECK13-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
// CHECK13-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP36]], align 8
// CHECK13-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP37]], align 8
// CHECK13-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK13-NEXT:    store i64 [[TMP29]], i64* [[TMP38]], align 8
// CHECK13-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK13-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK13-NEXT:    br i1 [[TMP40]], 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_l148(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK13:       omp_offload.cont:
// CHECK13-NEXT:    [[TMP41:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
// CHECK13-NEXT:    store i32 [[TMP41]], i32* [[CONV4]], align 4
// CHECK13-NEXT:    [[TMP42:%.*]] = load i64, i64* [[N_CASTED3]], align 8
// CHECK13-NEXT:    [[TMP43:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP44:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES8]] to i8*
// CHECK13-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP44]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes.4 to i8*), i64 24, i1 false)
// CHECK13-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK13-NEXT:    store i64 [[TMP42]], i64* [[TMP46]], align 8
// CHECK13-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK13-NEXT:    store i64 [[TMP42]], i64* [[TMP48]], align 8
// CHECK13-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK13-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP51]], align 8
// CHECK13-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP53]], align 8
// CHECK13-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK13-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP56]], align 8
// CHECK13-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP58]], align 8
// CHECK13-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
// CHECK13-NEXT:    store i64 [[TMP43]], i64* [[TMP59]], align 8
// CHECK13-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK13-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP64:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP64]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK13-NEXT:    [[TMP65:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK13-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP65]], 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:    [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK13-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP66]], 1
// CHECK13-NEXT:    [[TMP67:%.*]] = zext i32 [[ADD15]] to i64
// CHECK13-NEXT:    [[KERNEL_ARGS16:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK13-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 1
// CHECK13-NEXT:    store i32 3, i32* [[TMP69]], align 4
// CHECK13-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP61]], i8*** [[TMP70]], align 8
// CHECK13-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP62]], i8*** [[TMP71]], align 8
// CHECK13-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 4
// CHECK13-NEXT:    store i64* [[TMP63]], i64** [[TMP72]], align 8
// CHECK13-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP73]], align 8
// CHECK13-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP74]], align 8
// CHECK13-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP75]], align 8
// CHECK13-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 8
// CHECK13-NEXT:    store i64 [[TMP67]], i64* [[TMP76]], align 8
// CHECK13-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]])
// CHECK13-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK13-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK13:       omp_offload.failed17:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i64 [[TMP42]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK13:       omp_offload.cont18:
// CHECK13-NEXT:    [[TMP79:%.*]] = load i32, i32* [[M]], align 4
// CHECK13-NEXT:    [[CONV19:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP79]], i32* [[CONV19]], align 4
// CHECK13-NEXT:    [[TMP80:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK13-NEXT:    [[TMP81:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV21:%.*]] = bitcast i64* [[N_CASTED20]] to i32*
// CHECK13-NEXT:    store i32 [[TMP81]], i32* [[CONV21]], align 4
// CHECK13-NEXT:    [[TMP82:%.*]] = load i64, i64* [[N_CASTED20]], align 8
// CHECK13-NEXT:    [[TMP83:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP84:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES25]] to i8*
// CHECK13-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP84]], i8* align 8 bitcast ([4 x i64]* @.offload_sizes.8 to i8*), i64 32, i1 false)
// CHECK13-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
// CHECK13-NEXT:    store i64 [[TMP80]], i64* [[TMP86]], align 8
// CHECK13-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
// CHECK13-NEXT:    store i64 [[TMP80]], i64* [[TMP88]], align 8
// CHECK13-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP89]], align 8
// CHECK13-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i64*
// CHECK13-NEXT:    store i64 [[TMP82]], i64* [[TMP91]], align 8
// CHECK13-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i64*
// CHECK13-NEXT:    store i64 [[TMP82]], i64* [[TMP93]], align 8
// CHECK13-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP94]], align 8
// CHECK13-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP96]], align 8
// CHECK13-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP98]], align 8
// CHECK13-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP99]], align 8
// CHECK13-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 3
// CHECK13-NEXT:    [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP101]], align 8
// CHECK13-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 3
// CHECK13-NEXT:    [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP103]], align 8
// CHECK13-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 3
// CHECK13-NEXT:    store i64 [[TMP83]], i64* [[TMP104]], align 8
// CHECK13-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 3
// CHECK13-NEXT:    store i8* null, i8** [[TMP105]], align 8
// CHECK13-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP109:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP109]], i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK13-NEXT:    [[TMP110:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK13-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[TMP110]], 0
// CHECK13-NEXT:    [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
// CHECK13-NEXT:    [[SUB31:%.*]] = sub nsw i32 [[DIV30]], 1
// CHECK13-NEXT:    store i32 [[SUB31]], i32* [[DOTCAPTURE_EXPR_28]], align 4
// CHECK13-NEXT:    [[TMP111:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_28]], align 4
// CHECK13-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP111]], 1
// CHECK13-NEXT:    [[TMP112:%.*]] = zext i32 [[ADD32]] to i64
// CHECK13-NEXT:    [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP113]], align 4
// CHECK13-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 1
// CHECK13-NEXT:    store i32 4, i32* [[TMP114]], align 4
// CHECK13-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP106]], i8*** [[TMP115]], align 8
// CHECK13-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP107]], i8*** [[TMP116]], align 8
// CHECK13-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 4
// CHECK13-NEXT:    store i64* [[TMP108]], i64** [[TMP117]], align 8
// CHECK13-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP118]], align 8
// CHECK13-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP119]], align 8
// CHECK13-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP120]], align 8
// CHECK13-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 8
// CHECK13-NEXT:    store i64 [[TMP112]], i64* [[TMP121]], align 8
// CHECK13-NEXT:    [[TMP122:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]])
// CHECK13-NEXT:    [[TMP123:%.*]] = icmp ne i32 [[TMP122]], 0
// CHECK13-NEXT:    br i1 [[TMP123]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
// CHECK13:       omp_offload.failed34:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i64 [[TMP80]], i64 [[TMP82]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT35]]
// CHECK13:       omp_offload.cont35:
// CHECK13-NEXT:    [[TMP124:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV37:%.*]] = bitcast i64* [[N_CASTED36]] to i32*
// CHECK13-NEXT:    store i32 [[TMP124]], i32* [[CONV37]], align 4
// CHECK13-NEXT:    [[TMP125:%.*]] = load i64, i64* [[N_CASTED36]], align 8
// CHECK13-NEXT:    [[TMP126:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP127:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES41]] to i8*
// CHECK13-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP127]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes.12 to i8*), i64 24, i1 false)
// CHECK13-NEXT:    [[TMP128:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP129:%.*]] = bitcast i8** [[TMP128]] to i64*
// CHECK13-NEXT:    store i64 [[TMP125]], i64* [[TMP129]], align 8
// CHECK13-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i64*
// CHECK13-NEXT:    store i64 [[TMP125]], i64* [[TMP131]], align 8
// CHECK13-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP132]], align 8
// CHECK13-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP134:%.*]] = bitcast i8** [[TMP133]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP134]], align 8
// CHECK13-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP136]], align 8
// CHECK13-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP137]], align 8
// CHECK13-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP139]], align 8
// CHECK13-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP141]], align 8
// CHECK13-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES41]], i32 0, i32 2
// CHECK13-NEXT:    store i64 [[TMP126]], i64* [[TMP142]], align 8
// CHECK13-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP143]], align 8
// CHECK13-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES41]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP147:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP147]], i32* [[DOTCAPTURE_EXPR_43]], align 4
// CHECK13-NEXT:    [[TMP148:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_43]], align 4
// CHECK13-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[TMP148]], 0
// CHECK13-NEXT:    [[DIV46:%.*]] = sdiv i32 [[SUB45]], 1
// CHECK13-NEXT:    [[SUB47:%.*]] = sub nsw i32 [[DIV46]], 1
// CHECK13-NEXT:    store i32 [[SUB47]], i32* [[DOTCAPTURE_EXPR_44]], align 4
// CHECK13-NEXT:    [[TMP149:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_44]], align 4
// CHECK13-NEXT:    [[ADD48:%.*]] = add nsw i32 [[TMP149]], 1
// CHECK13-NEXT:    [[TMP150:%.*]] = zext i32 [[ADD48]] to i64
// CHECK13-NEXT:    [[KERNEL_ARGS49:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP151]], align 4
// CHECK13-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 1
// CHECK13-NEXT:    store i32 3, i32* [[TMP152]], align 4
// CHECK13-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP144]], i8*** [[TMP153]], align 8
// CHECK13-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP145]], i8*** [[TMP154]], align 8
// CHECK13-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 4
// CHECK13-NEXT:    store i64* [[TMP146]], i64** [[TMP155]], align 8
// CHECK13-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP156]], align 8
// CHECK13-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP157]], align 8
// CHECK13-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP158]], align 8
// CHECK13-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 8
// CHECK13-NEXT:    store i64 [[TMP150]], i64* [[TMP159]], align 8
// CHECK13-NEXT:    [[TMP160:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]])
// CHECK13-NEXT:    [[TMP161:%.*]] = icmp ne i32 [[TMP160]], 0
// CHECK13-NEXT:    br i1 [[TMP161]], label [[OMP_OFFLOAD_FAILED50:%.*]], label [[OMP_OFFLOAD_CONT51:%.*]]
// CHECK13:       omp_offload.failed50:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i64 [[TMP125]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT51]]
// CHECK13:       omp_offload.cont51:
// CHECK13-NEXT:    [[TMP162:%.*]] = load i32, i32* [[M]], align 4
// CHECK13-NEXT:    [[CONV53:%.*]] = bitcast i64* [[M_CASTED52]] to i32*
// CHECK13-NEXT:    store i32 [[TMP162]], i32* [[CONV53]], align 4
// CHECK13-NEXT:    [[TMP163:%.*]] = load i64, i64* [[M_CASTED52]], align 8
// CHECK13-NEXT:    [[TMP164:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    [[CONV55:%.*]] = bitcast i64* [[N_CASTED54]] to i32*
// CHECK13-NEXT:    store i32 [[TMP164]], i32* [[CONV55]], align 4
// CHECK13-NEXT:    [[TMP165:%.*]] = load i64, i64* [[N_CASTED54]], align 8
// CHECK13-NEXT:    [[TMP166:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK13-NEXT:    [[TMP167:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES59]] to i8*
// CHECK13-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP167]], i8* align 8 bitcast ([4 x i64]* @.offload_sizes.16 to i8*), i64 32, i1 false)
// CHECK13-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP169:%.*]] = bitcast i8** [[TMP168]] to i64*
// CHECK13-NEXT:    store i64 [[TMP163]], i64* [[TMP169]], align 8
// CHECK13-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP171:%.*]] = bitcast i8** [[TMP170]] to i64*
// CHECK13-NEXT:    store i64 [[TMP163]], i64* [[TMP171]], align 8
// CHECK13-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP172]], align 8
// CHECK13-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP174:%.*]] = bitcast i8** [[TMP173]] to i64*
// CHECK13-NEXT:    store i64 [[TMP165]], i64* [[TMP174]], align 8
// CHECK13-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP176:%.*]] = bitcast i8** [[TMP175]] to i64*
// CHECK13-NEXT:    store i64 [[TMP165]], i64* [[TMP176]], align 8
// CHECK13-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP177]], align 8
// CHECK13-NEXT:    [[TMP178:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP179:%.*]] = bitcast i8** [[TMP178]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP179]], align 8
// CHECK13-NEXT:    [[TMP180:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 2
// CHECK13-NEXT:    [[TMP181:%.*]] = bitcast i8** [[TMP180]] to i64*
// CHECK13-NEXT:    store i64 [[TMP1]], i64* [[TMP181]], align 8
// CHECK13-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 2
// CHECK13-NEXT:    store i8* null, i8** [[TMP182]], align 8
// CHECK13-NEXT:    [[TMP183:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 3
// CHECK13-NEXT:    [[TMP184:%.*]] = bitcast i8** [[TMP183]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP184]], align 8
// CHECK13-NEXT:    [[TMP185:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 3
// CHECK13-NEXT:    [[TMP186:%.*]] = bitcast i8** [[TMP185]] to i32**
// CHECK13-NEXT:    store i32* [[VLA]], i32** [[TMP186]], align 8
// CHECK13-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES59]], i32 0, i32 3
// CHECK13-NEXT:    store i64 [[TMP166]], i64* [[TMP187]], align 8
// CHECK13-NEXT:    [[TMP188:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 3
// CHECK13-NEXT:    store i8* null, i8** [[TMP188]], align 8
// CHECK13-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP190:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES59]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP192:%.*]] = load i32, i32* [[N]], align 4
// CHECK13-NEXT:    store i32 [[TMP192]], i32* [[DOTCAPTURE_EXPR_61]], align 4
// CHECK13-NEXT:    [[TMP193:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_61]], align 4
// CHECK13-NEXT:    [[SUB63:%.*]] = sub nsw i32 [[TMP193]], 0
// CHECK13-NEXT:    [[DIV64:%.*]] = sdiv i32 [[SUB63]], 1
// CHECK13-NEXT:    [[SUB65:%.*]] = sub nsw i32 [[DIV64]], 1
// CHECK13-NEXT:    store i32 [[SUB65]], i32* [[DOTCAPTURE_EXPR_62]], align 4
// CHECK13-NEXT:    [[TMP194:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_62]], align 4
// CHECK13-NEXT:    [[ADD66:%.*]] = add nsw i32 [[TMP194]], 1
// CHECK13-NEXT:    [[TMP195:%.*]] = zext i32 [[ADD66]] to i64
// CHECK13-NEXT:    [[KERNEL_ARGS67:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP196:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP196]], align 4
// CHECK13-NEXT:    [[TMP197:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 1
// CHECK13-NEXT:    store i32 4, i32* [[TMP197]], align 4
// CHECK13-NEXT:    [[TMP198:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP189]], i8*** [[TMP198]], align 8
// CHECK13-NEXT:    [[TMP199:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP190]], i8*** [[TMP199]], align 8
// CHECK13-NEXT:    [[TMP200:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 4
// CHECK13-NEXT:    store i64* [[TMP191]], i64** [[TMP200]], align 8
// CHECK13-NEXT:    [[TMP201:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP201]], align 8
// CHECK13-NEXT:    [[TMP202:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP202]], align 8
// CHECK13-NEXT:    [[TMP203:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP203]], align 8
// CHECK13-NEXT:    [[TMP204:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 8
// CHECK13-NEXT:    store i64 [[TMP195]], i64* [[TMP204]], align 8
// CHECK13-NEXT:    [[TMP205:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]])
// CHECK13-NEXT:    [[TMP206:%.*]] = icmp ne i32 [[TMP205]], 0
// CHECK13-NEXT:    br i1 [[TMP206]], label [[OMP_OFFLOAD_FAILED68:%.*]], label [[OMP_OFFLOAD_CONT69:%.*]]
// CHECK13:       omp_offload.failed68:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i64 [[TMP163]], i64 [[TMP165]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT69]]
// CHECK13:       omp_offload.cont69:
// CHECK13-NEXT:    [[TMP207:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK13-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP207]])
// CHECK13-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK13-NEXT:    [[TMP208:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK13-NEXT:    call void @llvm.stackrestore(i8* [[TMP208]])
// CHECK13-NEXT:    [[TMP209:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK13-NEXT:    ret i32 [[TMP209]]
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
// CHECK13-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
// 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
// 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
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    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_l153
// CHECK13-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
// 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
// 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
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    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_l158
// CHECK13-SAME: (i64 noundef [[M:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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 4
// 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..6 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[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 4
// 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
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// 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
// CHECK13-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK13-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP21]], i32* [[CONV7]], align 4
// CHECK13-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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..7 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// 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
// CHECK13-NEXT:    br label [[COND_END14:%.*]]
// CHECK13:       cond.false13:
// CHECK13-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[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
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
// 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
// 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
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    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_l163
// CHECK13-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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..10 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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..11 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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 1073741859, 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 !15
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
// 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 !15
// 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 !15
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !15
// 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 !15
// 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 !15
// CHECK13-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK13-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![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:    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_l168
// CHECK13-SAME: (i64 noundef [[M:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef 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 4
// 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..14 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[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
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK13-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP20]], i32* [[CONV7]], align 4
// CHECK13-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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..15 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[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 4
// 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 1073741859, 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 !18
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// 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 !18
// 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 !18
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !18
// 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 !18
// 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 !18
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![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:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK13-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[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_BASEPTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [1 x i8*], align 8
// CHECK13-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[M_CASTED22:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [2 x i8*], align 8
// CHECK13-NEXT:    [[_TMP27:%.*]] = 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK13-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK13-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK13-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK13-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 8
// CHECK13-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK13-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP11]], align 8
// CHECK13-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP12]], align 8
// CHECK13-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP13]], align 8
// CHECK13-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP14]], align 8
// CHECK13-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK13-NEXT:    store i64 10, i64* [[TMP15]], align 8
// CHECK13-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK13-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TMP17]], 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__l116([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK13:       omp_offload.cont:
// CHECK13-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 8
// CHECK13-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK13-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK13-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK13-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK13-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 8
// CHECK13-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 8
// CHECK13-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK13-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.24, i32 0, i32 0), i64** [[TMP29]], align 8
// CHECK13-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.25, i32 0, i32 0), i64** [[TMP30]], align 8
// CHECK13-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP31]], align 8
// CHECK13-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP32]], align 8
// CHECK13-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK13-NEXT:    store i64 10, i64* [[TMP33]], align 8
// CHECK13-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
// CHECK13-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK13-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK13:       omp_offload.failed6:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK13:       omp_offload.cont7:
// CHECK13-NEXT:    [[TMP36:%.*]] = load i32, i32* [[M]], align 4
// CHECK13-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP36]], i32* [[CONV]], align 4
// CHECK13-NEXT:    [[TMP37:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK13-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK13-NEXT:    store i64 [[TMP37]], i64* [[TMP39]], align 8
// CHECK13-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
// CHECK13-NEXT:    store i64 [[TMP37]], i64* [[TMP41]], align 8
// CHECK13-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP42]], align 8
// CHECK13-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 8
// CHECK13-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 8
// CHECK13-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP47]], align 8
// CHECK13-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK13-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP50]], align 4
// CHECK13-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
// CHECK13-NEXT:    store i32 2, i32* [[TMP51]], align 4
// CHECK13-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP48]], i8*** [[TMP52]], align 8
// CHECK13-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP49]], i8*** [[TMP53]], align 8
// CHECK13-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
// CHECK13-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.28, i32 0, i32 0), i64** [[TMP54]], align 8
// CHECK13-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.29, i32 0, i32 0), i64** [[TMP55]], align 8
// CHECK13-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP56]], align 8
// CHECK13-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP57]], align 8
// CHECK13-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
// CHECK13-NEXT:    store i64 10, i64* [[TMP58]], align 8
// CHECK13-NEXT:    [[TMP59:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
// CHECK13-NEXT:    [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
// CHECK13-NEXT:    br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
// CHECK13:       omp_offload.failed13:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i64 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
// CHECK13:       omp_offload.cont14:
// CHECK13-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP62]], align 8
// CHECK13-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP64]], align 8
// CHECK13-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS17]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP65]], align 8
// CHECK13-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK13-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK13-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 1
// CHECK13-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK13-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP66]], i8*** [[TMP70]], align 8
// CHECK13-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP67]], i8*** [[TMP71]], align 8
// CHECK13-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 4
// CHECK13-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.32, i32 0, i32 0), i64** [[TMP72]], align 8
// CHECK13-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.33, i32 0, i32 0), i64** [[TMP73]], align 8
// CHECK13-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP74]], align 8
// CHECK13-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP75]], align 8
// CHECK13-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 8
// CHECK13-NEXT:    store i64 10, i64* [[TMP76]], align 8
// CHECK13-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]])
// CHECK13-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK13-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
// CHECK13:       omp_offload.failed20:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
// CHECK13:       omp_offload.cont21:
// CHECK13-NEXT:    [[TMP79:%.*]] = load i32, i32* [[M]], align 4
// CHECK13-NEXT:    [[CONV23:%.*]] = bitcast i64* [[M_CASTED22]] to i32*
// CHECK13-NEXT:    store i32 [[TMP79]], i32* [[CONV23]], align 4
// CHECK13-NEXT:    [[TMP80:%.*]] = load i64, i64* [[M_CASTED22]], align 8
// CHECK13-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
// CHECK13-NEXT:    store i64 [[TMP80]], i64* [[TMP82]], align 8
// CHECK13-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP84:%.*]] = bitcast i8** [[TMP83]] to i64*
// CHECK13-NEXT:    store i64 [[TMP80]], i64* [[TMP84]], align 8
// CHECK13-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
// CHECK13-NEXT:    store i8* null, i8** [[TMP85]], align 8
// CHECK13-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP87]], align 8
// CHECK13-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 1
// CHECK13-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to [10 x i32]**
// CHECK13-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP89]], align 8
// CHECK13-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 1
// CHECK13-NEXT:    store i8* null, i8** [[TMP90]], align 8
// CHECK13-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK13-NEXT:    [[KERNEL_ARGS28:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 0
// CHECK13-NEXT:    store i32 1, i32* [[TMP93]], align 4
// CHECK13-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 1
// CHECK13-NEXT:    store i32 2, i32* [[TMP94]], align 4
// CHECK13-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 2
// CHECK13-NEXT:    store i8** [[TMP91]], i8*** [[TMP95]], align 8
// CHECK13-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 3
// CHECK13-NEXT:    store i8** [[TMP92]], i8*** [[TMP96]], align 8
// CHECK13-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 4
// CHECK13-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.36, i32 0, i32 0), i64** [[TMP97]], align 8
// CHECK13-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 5
// CHECK13-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.37, i32 0, i32 0), i64** [[TMP98]], align 8
// CHECK13-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 6
// CHECK13-NEXT:    store i8** null, i8*** [[TMP99]], align 8
// CHECK13-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 7
// CHECK13-NEXT:    store i8** null, i8*** [[TMP100]], align 8
// CHECK13-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 8
// CHECK13-NEXT:    store i64 10, i64* [[TMP101]], align 8
// CHECK13-NEXT:    [[TMP102:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]])
// CHECK13-NEXT:    [[TMP103:%.*]] = icmp ne i32 [[TMP102]], 0
// CHECK13-NEXT:    br i1 [[TMP103]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
// CHECK13:       omp_offload.failed29:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i64 [[TMP80]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT30]]
// CHECK13:       omp_offload.cont30:
// CHECK13-NEXT:    ret i32 0
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
// CHECK13-SAME: ([10 x i32]* noundef 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..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef 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
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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..19 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..19
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef 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
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
// CHECK13-SAME: ([10 x i32]* noundef 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..22 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef 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
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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..23 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..23
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef 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
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
// CHECK13-SAME: (i64 noundef [[M:%.*]], [10 x i32]* noundef 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 4
// 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..26 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[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
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK13-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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..27 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..27
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[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 4
// 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
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
// CHECK13-SAME: ([10 x i32]* noundef 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..30 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef 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
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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..31 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..31
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef 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 1073741859, 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 !21
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
// 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 !21
// 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 !21
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
// 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 !21
// 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 !21
// CHECK13-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK13-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
// CHECK13-SAME: (i64 noundef [[M:%.*]], [10 x i32]* noundef 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 4
// 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..34 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..34
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[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
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK13-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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..35 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..35
// CHECK13-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[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 4
// 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 1073741859, 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 !24
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
// 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 !24
// 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 !24
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !24
// 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 !24
// 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 !24
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![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:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK13-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK13-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@main
// CHECK15-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[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 4
// CHECK15-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK15-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], 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:    [[N_CASTED3:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
// CHECK15-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_CASTED18:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [4 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [4 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [4 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES22:%.*]] = alloca [4 x i64], align 4
// CHECK15-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_CASTED33:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS34:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS35:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS36:%.*]] = alloca [3 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES37:%.*]] = alloca [3 x i64], align 4
// CHECK15-NEXT:    [[_TMP38:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M_CASTED48:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_CASTED49:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS50:%.*]] = alloca [4 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS51:%.*]] = alloca [4 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS52:%.*]] = alloca [4 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_SIZES53:%.*]] = alloca [4 x i64], align 4
// CHECK15-NEXT:    [[_TMP54:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_55:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_56:%.*]] = 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 4
// CHECK15-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK15-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK15-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK15-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[N_CASTED]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK15-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
// CHECK15-NEXT:    [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK15-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
// CHECK15-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[TMP8]], align 4
// CHECK15-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[TMP10]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK15-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP13]], align 4
// CHECK15-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP15]], align 4
// CHECK15-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK15-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP18]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP20]], align 4
// CHECK15-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK15-NEXT:    store i64 [[TMP5]], i64* [[TMP21]], align 4
// CHECK15-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK15-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP27]], 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:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK15-NEXT:    [[TMP29:%.*]] = zext i32 [[ADD]] to i64
// CHECK15-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK15-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP30]], align 4
// CHECK15-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK15-NEXT:    store i32 3, i32* [[TMP31]], align 4
// CHECK15-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP23]], i8*** [[TMP32]], align 4
// CHECK15-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP24]], i8*** [[TMP33]], align 4
// CHECK15-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK15-NEXT:    store i64* [[TMP25]], i64** [[TMP34]], align 4
// CHECK15-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
// CHECK15-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP36]], align 4
// CHECK15-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP37]], align 4
// CHECK15-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK15-NEXT:    store i64 [[TMP29]], i64* [[TMP38]], align 8
// CHECK15-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK15-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK15-NEXT:    br i1 [[TMP40]], 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_l148(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK15:       omp_offload.cont:
// CHECK15-NEXT:    [[TMP41:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP41]], i32* [[N_CASTED3]], align 4
// CHECK15-NEXT:    [[TMP42:%.*]] = load i32, i32* [[N_CASTED3]], align 4
// CHECK15-NEXT:    [[TMP43:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK15-NEXT:    [[TMP44:%.*]] = sext i32 [[TMP43]] to i64
// CHECK15-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES7]] to i8*
// CHECK15-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.4 to i8*), i32 24, i1 false)
// CHECK15-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
// CHECK15-NEXT:    store i32 [[TMP42]], i32* [[TMP47]], align 4
// CHECK15-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
// CHECK15-NEXT:    store i32 [[TMP42]], i32* [[TMP49]], align 4
// CHECK15-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP50]], align 4
// CHECK15-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP52]], align 4
// CHECK15-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP54:%.*]] = bitcast i8** [[TMP53]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP54]], align 4
// CHECK15-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP55]], align 4
// CHECK15-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP57]], align 4
// CHECK15-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP59]], align 4
// CHECK15-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
// CHECK15-NEXT:    store i64 [[TMP44]], i64* [[TMP60]], align 4
// CHECK15-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP61]], align 4
// CHECK15-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP65:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP65]], i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK15-NEXT:    [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK15-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP66]], 0
// CHECK15-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
// CHECK15-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
// CHECK15-NEXT:    store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK15-NEXT:    [[TMP67:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK15-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP67]], 1
// CHECK15-NEXT:    [[TMP68:%.*]] = zext i32 [[ADD14]] to i64
// CHECK15-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK15-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK15-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK15-NEXT:    store i32 3, i32* [[TMP70]], align 4
// CHECK15-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP62]], i8*** [[TMP71]], align 4
// CHECK15-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP63]], i8*** [[TMP72]], align 4
// CHECK15-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK15-NEXT:    store i64* [[TMP64]], i64** [[TMP73]], align 4
// CHECK15-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP74]], align 4
// CHECK15-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP75]], align 4
// CHECK15-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP76]], align 4
// CHECK15-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK15-NEXT:    store i64 [[TMP68]], i64* [[TMP77]], align 8
// CHECK15-NEXT:    [[TMP78:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK15-NEXT:    [[TMP79:%.*]] = icmp ne i32 [[TMP78]], 0
// CHECK15-NEXT:    br i1 [[TMP79]], 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_l153(i32 [[TMP42]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK15:       omp_offload.cont17:
// CHECK15-NEXT:    [[TMP80:%.*]] = load i32, i32* [[M]], align 4
// CHECK15-NEXT:    store i32 [[TMP80]], i32* [[M_CASTED]], align 4
// CHECK15-NEXT:    [[TMP81:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK15-NEXT:    [[TMP82:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP82]], i32* [[N_CASTED18]], align 4
// CHECK15-NEXT:    [[TMP83:%.*]] = load i32, i32* [[N_CASTED18]], align 4
// CHECK15-NEXT:    [[TMP84:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK15-NEXT:    [[TMP85:%.*]] = sext i32 [[TMP84]] to i64
// CHECK15-NEXT:    [[TMP86:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES22]] to i8*
// CHECK15-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP86]], i8* align 4 bitcast ([4 x i64]* @.offload_sizes.8 to i8*), i32 32, i1 false)
// CHECK15-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
// CHECK15-NEXT:    store i32 [[TMP81]], i32* [[TMP88]], align 4
// CHECK15-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP90:%.*]] = bitcast i8** [[TMP89]] to i32*
// CHECK15-NEXT:    store i32 [[TMP81]], i32* [[TMP90]], align 4
// CHECK15-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP91]], align 4
// CHECK15-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i32*
// CHECK15-NEXT:    store i32 [[TMP83]], i32* [[TMP93]], align 4
// CHECK15-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP95:%.*]] = bitcast i8** [[TMP94]] to i32*
// CHECK15-NEXT:    store i32 [[TMP83]], i32* [[TMP95]], align 4
// CHECK15-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP96]], align 4
// CHECK15-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP98]], align 4
// CHECK15-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP100]], align 4
// CHECK15-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP101]], align 4
// CHECK15-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 3
// CHECK15-NEXT:    [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP103]], align 4
// CHECK15-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 3
// CHECK15-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP105]], align 4
// CHECK15-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 3
// CHECK15-NEXT:    store i64 [[TMP85]], i64* [[TMP106]], align 4
// CHECK15-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 3
// CHECK15-NEXT:    store i8* null, i8** [[TMP107]], align 4
// CHECK15-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP111:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP111]], i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK15-NEXT:    [[TMP112:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK15-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[TMP112]], 0
// CHECK15-NEXT:    [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
// CHECK15-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[DIV27]], 1
// CHECK15-NEXT:    store i32 [[SUB28]], i32* [[DOTCAPTURE_EXPR_25]], align 4
// CHECK15-NEXT:    [[TMP113:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
// CHECK15-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP113]], 1
// CHECK15-NEXT:    [[TMP114:%.*]] = zext i32 [[ADD29]] to i64
// CHECK15-NEXT:    [[KERNEL_ARGS30:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK15-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP115]], align 4
// CHECK15-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 1
// CHECK15-NEXT:    store i32 4, i32* [[TMP116]], align 4
// CHECK15-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP108]], i8*** [[TMP117]], align 4
// CHECK15-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP109]], i8*** [[TMP118]], align 4
// CHECK15-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 4
// CHECK15-NEXT:    store i64* [[TMP110]], i64** [[TMP119]], align 4
// CHECK15-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP120]], align 4
// CHECK15-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP121]], align 4
// CHECK15-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP122]], align 4
// CHECK15-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 8
// CHECK15-NEXT:    store i64 [[TMP114]], i64* [[TMP123]], align 8
// CHECK15-NEXT:    [[TMP124:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]])
// CHECK15-NEXT:    [[TMP125:%.*]] = icmp ne i32 [[TMP124]], 0
// CHECK15-NEXT:    br i1 [[TMP125]], label [[OMP_OFFLOAD_FAILED31:%.*]], label [[OMP_OFFLOAD_CONT32:%.*]]
// CHECK15:       omp_offload.failed31:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i32 [[TMP81]], i32 [[TMP83]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT32]]
// CHECK15:       omp_offload.cont32:
// CHECK15-NEXT:    [[TMP126:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP126]], i32* [[N_CASTED33]], align 4
// CHECK15-NEXT:    [[TMP127:%.*]] = load i32, i32* [[N_CASTED33]], align 4
// CHECK15-NEXT:    [[TMP128:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK15-NEXT:    [[TMP129:%.*]] = sext i32 [[TMP128]] to i64
// CHECK15-NEXT:    [[TMP130:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES37]] to i8*
// CHECK15-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP130]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.12 to i8*), i32 24, i1 false)
// CHECK15-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP132:%.*]] = bitcast i8** [[TMP131]] to i32*
// CHECK15-NEXT:    store i32 [[TMP127]], i32* [[TMP132]], align 4
// CHECK15-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP134:%.*]] = bitcast i8** [[TMP133]] to i32*
// CHECK15-NEXT:    store i32 [[TMP127]], i32* [[TMP134]], align 4
// CHECK15-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP135]], align 4
// CHECK15-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP137]], align 4
// CHECK15-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP139]], align 4
// CHECK15-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP140]], align 4
// CHECK15-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP142]], align 4
// CHECK15-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP144]], align 4
// CHECK15-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES37]], i32 0, i32 2
// CHECK15-NEXT:    store i64 [[TMP129]], i64* [[TMP145]], align 4
// CHECK15-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP146]], align 4
// CHECK15-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP149:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES37]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP150:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP150]], i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK15-NEXT:    [[TMP151:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK15-NEXT:    [[SUB41:%.*]] = sub nsw i32 [[TMP151]], 0
// CHECK15-NEXT:    [[DIV42:%.*]] = sdiv i32 [[SUB41]], 1
// CHECK15-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[DIV42]], 1
// CHECK15-NEXT:    store i32 [[SUB43]], i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK15-NEXT:    [[TMP152:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK15-NEXT:    [[ADD44:%.*]] = add nsw i32 [[TMP152]], 1
// CHECK15-NEXT:    [[TMP153:%.*]] = zext i32 [[ADD44]] to i64
// CHECK15-NEXT:    [[KERNEL_ARGS45:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK15-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP154]], align 4
// CHECK15-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 1
// CHECK15-NEXT:    store i32 3, i32* [[TMP155]], align 4
// CHECK15-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP147]], i8*** [[TMP156]], align 4
// CHECK15-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP148]], i8*** [[TMP157]], align 4
// CHECK15-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 4
// CHECK15-NEXT:    store i64* [[TMP149]], i64** [[TMP158]], align 4
// CHECK15-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP159]], align 4
// CHECK15-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP160]], align 4
// CHECK15-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP161]], align 4
// CHECK15-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 8
// CHECK15-NEXT:    store i64 [[TMP153]], i64* [[TMP162]], align 8
// CHECK15-NEXT:    [[TMP163:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]])
// CHECK15-NEXT:    [[TMP164:%.*]] = icmp ne i32 [[TMP163]], 0
// CHECK15-NEXT:    br i1 [[TMP164]], label [[OMP_OFFLOAD_FAILED46:%.*]], label [[OMP_OFFLOAD_CONT47:%.*]]
// CHECK15:       omp_offload.failed46:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i32 [[TMP127]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT47]]
// CHECK15:       omp_offload.cont47:
// CHECK15-NEXT:    [[TMP165:%.*]] = load i32, i32* [[M]], align 4
// CHECK15-NEXT:    store i32 [[TMP165]], i32* [[M_CASTED48]], align 4
// CHECK15-NEXT:    [[TMP166:%.*]] = load i32, i32* [[M_CASTED48]], align 4
// CHECK15-NEXT:    [[TMP167:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP167]], i32* [[N_CASTED49]], align 4
// CHECK15-NEXT:    [[TMP168:%.*]] = load i32, i32* [[N_CASTED49]], align 4
// CHECK15-NEXT:    [[TMP169:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK15-NEXT:    [[TMP170:%.*]] = sext i32 [[TMP169]] to i64
// CHECK15-NEXT:    [[TMP171:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES53]] to i8*
// CHECK15-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP171]], i8* align 4 bitcast ([4 x i64]* @.offload_sizes.16 to i8*), i32 32, i1 false)
// CHECK15-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP173:%.*]] = bitcast i8** [[TMP172]] to i32*
// CHECK15-NEXT:    store i32 [[TMP166]], i32* [[TMP173]], align 4
// CHECK15-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP175:%.*]] = bitcast i8** [[TMP174]] to i32*
// CHECK15-NEXT:    store i32 [[TMP166]], i32* [[TMP175]], align 4
// CHECK15-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP176]], align 4
// CHECK15-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP178:%.*]] = bitcast i8** [[TMP177]] to i32*
// CHECK15-NEXT:    store i32 [[TMP168]], i32* [[TMP178]], align 4
// CHECK15-NEXT:    [[TMP179:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP180:%.*]] = bitcast i8** [[TMP179]] to i32*
// CHECK15-NEXT:    store i32 [[TMP168]], i32* [[TMP180]], align 4
// CHECK15-NEXT:    [[TMP181:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP181]], align 4
// CHECK15-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP183:%.*]] = bitcast i8** [[TMP182]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP183]], align 4
// CHECK15-NEXT:    [[TMP184:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 2
// CHECK15-NEXT:    [[TMP185:%.*]] = bitcast i8** [[TMP184]] to i32*
// CHECK15-NEXT:    store i32 [[TMP0]], i32* [[TMP185]], align 4
// CHECK15-NEXT:    [[TMP186:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 2
// CHECK15-NEXT:    store i8* null, i8** [[TMP186]], align 4
// CHECK15-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 3
// CHECK15-NEXT:    [[TMP188:%.*]] = bitcast i8** [[TMP187]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP188]], align 4
// CHECK15-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 3
// CHECK15-NEXT:    [[TMP190:%.*]] = bitcast i8** [[TMP189]] to i32**
// CHECK15-NEXT:    store i32* [[VLA]], i32** [[TMP190]], align 4
// CHECK15-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES53]], i32 0, i32 3
// CHECK15-NEXT:    store i64 [[TMP170]], i64* [[TMP191]], align 4
// CHECK15-NEXT:    [[TMP192:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 3
// CHECK15-NEXT:    store i8* null, i8** [[TMP192]], align 4
// CHECK15-NEXT:    [[TMP193:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP194:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP195:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES53]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP196:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    store i32 [[TMP196]], i32* [[DOTCAPTURE_EXPR_55]], align 4
// CHECK15-NEXT:    [[TMP197:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_55]], align 4
// CHECK15-NEXT:    [[SUB57:%.*]] = sub nsw i32 [[TMP197]], 0
// CHECK15-NEXT:    [[DIV58:%.*]] = sdiv i32 [[SUB57]], 1
// CHECK15-NEXT:    [[SUB59:%.*]] = sub nsw i32 [[DIV58]], 1
// CHECK15-NEXT:    store i32 [[SUB59]], i32* [[DOTCAPTURE_EXPR_56]], align 4
// CHECK15-NEXT:    [[TMP198:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_56]], align 4
// CHECK15-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP198]], 1
// CHECK15-NEXT:    [[TMP199:%.*]] = zext i32 [[ADD60]] to i64
// CHECK15-NEXT:    [[KERNEL_ARGS61:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK15-NEXT:    [[TMP200:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP200]], align 4
// CHECK15-NEXT:    [[TMP201:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 1
// CHECK15-NEXT:    store i32 4, i32* [[TMP201]], align 4
// CHECK15-NEXT:    [[TMP202:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP193]], i8*** [[TMP202]], align 4
// CHECK15-NEXT:    [[TMP203:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP194]], i8*** [[TMP203]], align 4
// CHECK15-NEXT:    [[TMP204:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 4
// CHECK15-NEXT:    store i64* [[TMP195]], i64** [[TMP204]], align 4
// CHECK15-NEXT:    [[TMP205:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP205]], align 4
// CHECK15-NEXT:    [[TMP206:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP206]], align 4
// CHECK15-NEXT:    [[TMP207:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP207]], align 4
// CHECK15-NEXT:    [[TMP208:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 8
// CHECK15-NEXT:    store i64 [[TMP199]], i64* [[TMP208]], align 8
// CHECK15-NEXT:    [[TMP209:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]])
// CHECK15-NEXT:    [[TMP210:%.*]] = icmp ne i32 [[TMP209]], 0
// CHECK15-NEXT:    br i1 [[TMP210]], label [[OMP_OFFLOAD_FAILED62:%.*]], label [[OMP_OFFLOAD_CONT63:%.*]]
// CHECK15:       omp_offload.failed62:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i32 [[TMP166]], i32 [[TMP168]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT63]]
// CHECK15:       omp_offload.cont63:
// CHECK15-NEXT:    [[TMP211:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK15-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP211]])
// CHECK15-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK15-NEXT:    [[TMP212:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK15-NEXT:    call void @llvm.stackrestore(i8* [[TMP212]])
// CHECK15-NEXT:    [[TMP213:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK15-NEXT:    ret i32 [[TMP213]]
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
// CHECK15-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// 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:    [[I3:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP8]], 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 4
// 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:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK15-NEXT:    br i1 [[CMP4]], 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
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK15-NEXT:    br i1 [[CMP5]], 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
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK15-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK15-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 4
// 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:    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_l153
// CHECK15-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// 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:    [[I3:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP8]], 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 4
// 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:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK15-NEXT:    br i1 [[CMP4]], 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
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK15-NEXT:    br i1 [[CMP5]], 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
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK15-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK15-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 4
// 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:    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_l158
// CHECK15-SAME: (i32 noundef [[M:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// 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 i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// 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
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// 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
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK15-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK15-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
// CHECK15-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK15:       cond.true11:
// CHECK15-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK15-NEXT:    br label [[COND_END13:%.*]]
// CHECK15:       cond.false12:
// CHECK15-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    br label [[COND_END13]]
// CHECK15:       cond.end13:
// CHECK15-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
// CHECK15-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP8]], 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 4
// 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:    [[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_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
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK15-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK15-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 4
// 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:    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_l163
// CHECK15-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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*, i32, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, i32*, i32, i32*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK15:       omp.inner.for.end:
// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK15:       omp.loop.exit:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK15-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// 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:    [[I3:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP8]], 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 4
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK15-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, 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 4
// 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 !16
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
// CHECK15-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK15-NEXT:    br i1 [[CMP4]], 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 !16
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !16
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
// 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 !16
// CHECK15-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK15-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![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:    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_l168
// CHECK15-SAME: (i32 noundef [[M:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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*, i32, i32*, i32)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// 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 i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 4
// 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:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// 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 i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP8]], 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* [[DOTCAPTURE_EXPR__ADDR]], align 4
// 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 4
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK15-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, 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 4
// 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 !19
// CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK15-NEXT:    br i1 [[CMP5]], 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 !19
// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !19
// CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !19
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
// 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 !19
// CHECK15-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK15-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![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:    br label [[OMP_PRECOND_END]]
// CHECK15:       omp.precond.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK15-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[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 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
// CHECK15-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 4
// CHECK15-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [1 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [1 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [1 x i8*], align 4
// CHECK15-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[M_CASTED22:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_BASEPTRS23:%.*]] = alloca [2 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_PTRS24:%.*]] = alloca [2 x i8*], align 4
// CHECK15-NEXT:    [[DOTOFFLOAD_MAPPERS25:%.*]] = alloca [2 x i8*], align 4
// CHECK15-NEXT:    [[_TMP26:%.*]] = 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 4
// 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 4
// CHECK15-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP4]], align 4
// 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK15-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK15-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK15-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK15-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK15-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP11]], align 4
// CHECK15-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP12]], align 4
// CHECK15-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP13]], align 4
// CHECK15-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP14]], align 4
// CHECK15-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK15-NEXT:    store i64 10, i64* [[TMP15]], align 8
// CHECK15-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK15-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK15-NEXT:    br i1 [[TMP17]], 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__l116([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK15:       omp_offload.cont:
// CHECK15-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 4
// CHECK15-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 4
// CHECK15-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK15-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK15-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK15-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK15-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK15-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK15-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 4
// CHECK15-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 4
// CHECK15-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK15-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.24, i32 0, i32 0), i64** [[TMP29]], align 4
// CHECK15-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.25, i32 0, i32 0), i64** [[TMP30]], align 4
// CHECK15-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP31]], align 4
// CHECK15-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP32]], align 4
// CHECK15-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK15-NEXT:    store i64 10, i64* [[TMP33]], align 8
// CHECK15-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
// CHECK15-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK15-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK15:       omp_offload.failed6:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK15:       omp_offload.cont7:
// CHECK15-NEXT:    [[TMP36:%.*]] = load i32, i32* [[M]], align 4
// CHECK15-NEXT:    store i32 [[TMP36]], i32* [[M_CASTED]], align 4
// CHECK15-NEXT:    [[TMP37:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK15-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i32*
// CHECK15-NEXT:    store i32 [[TMP37]], i32* [[TMP39]], align 4
// CHECK15-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i32*
// CHECK15-NEXT:    store i32 [[TMP37]], i32* [[TMP41]], align 4
// CHECK15-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP42]], align 4
// CHECK15-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 4
// CHECK15-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 4
// CHECK15-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP47]], align 4
// CHECK15-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK15-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK15-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP50]], align 4
// CHECK15-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
// CHECK15-NEXT:    store i32 2, i32* [[TMP51]], align 4
// CHECK15-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP48]], i8*** [[TMP52]], align 4
// CHECK15-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP49]], i8*** [[TMP53]], align 4
// CHECK15-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
// CHECK15-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.28, i32 0, i32 0), i64** [[TMP54]], align 4
// CHECK15-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.29, i32 0, i32 0), i64** [[TMP55]], align 4
// CHECK15-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP56]], align 4
// CHECK15-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP57]], align 4
// CHECK15-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
// CHECK15-NEXT:    store i64 10, i64* [[TMP58]], align 8
// CHECK15-NEXT:    [[TMP59:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
// CHECK15-NEXT:    [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
// CHECK15-NEXT:    br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
// CHECK15:       omp_offload.failed13:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i32 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
// CHECK15:       omp_offload.cont14:
// CHECK15-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP62]], align 4
// CHECK15-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP64]], align 4
// CHECK15-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS17]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP65]], align 4
// CHECK15-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK15-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK15-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK15-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 1
// CHECK15-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK15-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP66]], i8*** [[TMP70]], align 4
// CHECK15-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP67]], i8*** [[TMP71]], align 4
// CHECK15-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 4
// CHECK15-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.32, i32 0, i32 0), i64** [[TMP72]], align 4
// CHECK15-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.33, i32 0, i32 0), i64** [[TMP73]], align 4
// CHECK15-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP74]], align 4
// CHECK15-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP75]], align 4
// CHECK15-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 8
// CHECK15-NEXT:    store i64 10, i64* [[TMP76]], align 8
// CHECK15-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]])
// CHECK15-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK15-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
// CHECK15:       omp_offload.failed20:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
// CHECK15:       omp_offload.cont21:
// CHECK15-NEXT:    [[TMP79:%.*]] = load i32, i32* [[M]], align 4
// CHECK15-NEXT:    store i32 [[TMP79]], i32* [[M_CASTED22]], align 4
// CHECK15-NEXT:    [[TMP80:%.*]] = load i32, i32* [[M_CASTED22]], align 4
// CHECK15-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
// CHECK15-NEXT:    store i32 [[TMP80]], i32* [[TMP82]], align 4
// CHECK15-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP84:%.*]] = bitcast i8** [[TMP83]] to i32*
// CHECK15-NEXT:    store i32 [[TMP80]], i32* [[TMP84]], align 4
// CHECK15-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 0
// CHECK15-NEXT:    store i8* null, i8** [[TMP85]], align 4
// CHECK15-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP87]], align 4
// CHECK15-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 1
// CHECK15-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to [10 x i32]**
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP89]], align 4
// CHECK15-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 1
// CHECK15-NEXT:    store i8* null, i8** [[TMP90]], align 4
// CHECK15-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
// CHECK15-NEXT:    [[KERNEL_ARGS27:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK15-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 0
// CHECK15-NEXT:    store i32 1, i32* [[TMP93]], align 4
// CHECK15-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 1
// CHECK15-NEXT:    store i32 2, i32* [[TMP94]], align 4
// CHECK15-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 2
// CHECK15-NEXT:    store i8** [[TMP91]], i8*** [[TMP95]], align 4
// CHECK15-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 3
// CHECK15-NEXT:    store i8** [[TMP92]], i8*** [[TMP96]], align 4
// CHECK15-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 4
// CHECK15-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.36, i32 0, i32 0), i64** [[TMP97]], align 4
// CHECK15-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 5
// CHECK15-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.37, i32 0, i32 0), i64** [[TMP98]], align 4
// CHECK15-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 6
// CHECK15-NEXT:    store i8** null, i8*** [[TMP99]], align 4
// CHECK15-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 7
// CHECK15-NEXT:    store i8** null, i8*** [[TMP100]], align 4
// CHECK15-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 8
// CHECK15-NEXT:    store i64 10, i64* [[TMP101]], align 8
// CHECK15-NEXT:    [[TMP102:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]])
// CHECK15-NEXT:    [[TMP103:%.*]] = icmp ne i32 [[TMP102]], 0
// CHECK15-NEXT:    br i1 [[TMP103]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK15:       omp_offload.failed28:
// CHECK15-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i32 [[TMP80]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK15-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK15:       omp_offload.cont29:
// CHECK15-NEXT:    ret i32 0
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
// CHECK15-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// 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..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, [10 x i32]*)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..19
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], 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 4
// 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
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK15-NEXT:    br i1 [[CMP1]], 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
// 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
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK15-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK15-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
// CHECK15-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// 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..22 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, [10 x i32]*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..23
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], 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 4
// 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
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK15-NEXT:    br i1 [[CMP1]], 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
// 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
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK15-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK15-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
// CHECK15-SAME: (i32 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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]*, i32)* @.omp_outlined..26 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// 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 i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..27 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..27
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], 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* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// 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 i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK15:       cond.true:
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// 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 [ [[TMP8]], [[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:    [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK15-NEXT:    br i1 [[CMP1]], 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
// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK15-NEXT:    br i1 [[CMP2]], 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
// 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
// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK15-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK15-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK15-NEXT:    store i32 [[ADD4]], 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:    [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK15-NEXT:    store i32 [[ADD5]], 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:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
// CHECK15-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// 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..30 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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*, i32, i32, [10 x i32]*)* @.omp_outlined..31 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..31
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], 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 4
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK15-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, 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 !22
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
// 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 !22
// 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 !22
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !22
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !22
// 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 !22
// CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK15-NEXT:    store i32 [[ADD1]], 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_DISPATCH_INC:%.*]]
// CHECK15:       omp.dispatch.inc:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK15:       omp.dispatch.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
// CHECK15-SAME: (i32 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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]*, i32)* @.omp_outlined..34 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..34
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// 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 i32, align 4
// CHECK15-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// 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 4
// 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
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..35 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK15:       omp.inner.for.inc:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK15-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..35
// CHECK15-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// 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 4
// CHECK15-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], 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* [[DOTCAPTURE_EXPR__ADDR]], align 4
// 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 4
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK15-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, 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 !25
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
// 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 !25
// 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 !25
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
// 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:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK15-NEXT:    store i32 [[ADD1]], 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_DISPATCH_INC:%.*]]
// CHECK15:       omp.dispatch.inc:
// CHECK15-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK15:       omp.dispatch.end:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK15-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK15-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@main
// CHECK17-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[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 8
// CHECK17-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
// 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 i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
// CHECK17-NEXT:    [[_TMP9:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_CASTED20:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS23:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES25:%.*]] = alloca [4 x i64], align 8
// CHECK17-NEXT:    [[_TMP26:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_28:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED36:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS38:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS39:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS40:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES41:%.*]] = alloca [3 x i64], align 8
// CHECK17-NEXT:    [[_TMP42:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_43:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_44:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED52:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_CASTED54:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS56:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS57:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS58:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES59:%.*]] = alloca [4 x i64], align 8
// CHECK17-NEXT:    [[_TMP60:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_61:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_62:%.*]] = 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 8
// CHECK17-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK17-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK17-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK17-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK17-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK17-NEXT:    [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
// CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK17-NEXT:    store i64 [[TMP4]], i64* [[TMP8]], align 8
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK17-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP13]], align 8
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP15]], align 8
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP18]], align 8
// CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP20]], align 8
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP21]], align 8
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP27]], 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:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK17-NEXT:    [[TMP29:%.*]] = zext i32 [[ADD]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP30]], align 4
// CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK17-NEXT:    store i32 3, i32* [[TMP31]], align 4
// CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP23]], i8*** [[TMP32]], align 8
// CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP24]], i8*** [[TMP33]], align 8
// CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP25]], i64** [[TMP34]], align 8
// CHECK17-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
// CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP36]], align 8
// CHECK17-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP37]], align 8
// CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP29]], i64* [[TMP38]], align 8
// CHECK17-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK17-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK17-NEXT:    br i1 [[TMP40]], 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_l148(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[TMP41:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
// CHECK17-NEXT:    store i32 [[TMP41]], i32* [[CONV4]], align 4
// CHECK17-NEXT:    [[TMP42:%.*]] = load i64, i64* [[N_CASTED3]], align 8
// CHECK17-NEXT:    [[TMP43:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK17-NEXT:    [[TMP44:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES8]] to i8*
// CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP44]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes.4 to i8*), i64 24, i1 false)
// CHECK17-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK17-NEXT:    store i64 [[TMP42]], i64* [[TMP46]], align 8
// CHECK17-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK17-NEXT:    store i64 [[TMP42]], i64* [[TMP48]], align 8
// CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK17-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP51]], align 8
// CHECK17-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP53]], align 8
// CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP56]], align 8
// CHECK17-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP58]], align 8
// CHECK17-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP43]], i64* [[TMP59]], align 8
// CHECK17-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK17-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP64:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP64]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK17-NEXT:    [[TMP65:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK17-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP65]], 0
// CHECK17-NEXT:    [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
// CHECK17-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
// CHECK17-NEXT:    store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK17-NEXT:    [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK17-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP66]], 1
// CHECK17-NEXT:    [[TMP67:%.*]] = zext i32 [[ADD15]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS16:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK17-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 1
// CHECK17-NEXT:    store i32 3, i32* [[TMP69]], align 4
// CHECK17-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP61]], i8*** [[TMP70]], align 8
// CHECK17-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP62]], i8*** [[TMP71]], align 8
// CHECK17-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP63]], i64** [[TMP72]], align 8
// CHECK17-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP73]], align 8
// CHECK17-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP74]], align 8
// CHECK17-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP75]], align 8
// CHECK17-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP67]], i64* [[TMP76]], align 8
// CHECK17-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]])
// CHECK17-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK17-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK17:       omp_offload.failed17:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i64 [[TMP42]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK17:       omp_offload.cont18:
// CHECK17-NEXT:    [[TMP79:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    [[CONV19:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP79]], i32* [[CONV19]], align 4
// CHECK17-NEXT:    [[TMP80:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK17-NEXT:    [[TMP81:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[CONV21:%.*]] = bitcast i64* [[N_CASTED20]] to i32*
// CHECK17-NEXT:    store i32 [[TMP81]], i32* [[CONV21]], align 4
// CHECK17-NEXT:    [[TMP82:%.*]] = load i64, i64* [[N_CASTED20]], align 8
// CHECK17-NEXT:    [[TMP83:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK17-NEXT:    [[TMP84:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES25]] to i8*
// CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP84]], i8* align 8 bitcast ([4 x i64]* @.offload_sizes.8 to i8*), i64 32, i1 false)
// CHECK17-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
// CHECK17-NEXT:    store i64 [[TMP80]], i64* [[TMP86]], align 8
// CHECK17-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
// CHECK17-NEXT:    store i64 [[TMP80]], i64* [[TMP88]], align 8
// CHECK17-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP89]], align 8
// CHECK17-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i64*
// CHECK17-NEXT:    store i64 [[TMP82]], i64* [[TMP91]], align 8
// CHECK17-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i64*
// CHECK17-NEXT:    store i64 [[TMP82]], i64* [[TMP93]], align 8
// CHECK17-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP94]], align 8
// CHECK17-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP96]], align 8
// CHECK17-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP98]], align 8
// CHECK17-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP99]], align 8
// CHECK17-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP101]], align 8
// CHECK17-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP103]], align 8
// CHECK17-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP83]], i64* [[TMP104]], align 8
// CHECK17-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP105]], align 8
// CHECK17-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP109:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP109]], i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK17-NEXT:    [[TMP110:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK17-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[TMP110]], 0
// CHECK17-NEXT:    [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
// CHECK17-NEXT:    [[SUB31:%.*]] = sub nsw i32 [[DIV30]], 1
// CHECK17-NEXT:    store i32 [[SUB31]], i32* [[DOTCAPTURE_EXPR_28]], align 4
// CHECK17-NEXT:    [[TMP111:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_28]], align 4
// CHECK17-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP111]], 1
// CHECK17-NEXT:    [[TMP112:%.*]] = zext i32 [[ADD32]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP113]], align 4
// CHECK17-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 1
// CHECK17-NEXT:    store i32 4, i32* [[TMP114]], align 4
// CHECK17-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP106]], i8*** [[TMP115]], align 8
// CHECK17-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP107]], i8*** [[TMP116]], align 8
// CHECK17-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP108]], i64** [[TMP117]], align 8
// CHECK17-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP118]], align 8
// CHECK17-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP119]], align 8
// CHECK17-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP120]], align 8
// CHECK17-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP112]], i64* [[TMP121]], align 8
// CHECK17-NEXT:    [[TMP122:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]])
// CHECK17-NEXT:    [[TMP123:%.*]] = icmp ne i32 [[TMP122]], 0
// CHECK17-NEXT:    br i1 [[TMP123]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
// CHECK17:       omp_offload.failed34:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i64 [[TMP80]], i64 [[TMP82]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT35]]
// CHECK17:       omp_offload.cont35:
// CHECK17-NEXT:    [[TMP124:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[CONV37:%.*]] = bitcast i64* [[N_CASTED36]] to i32*
// CHECK17-NEXT:    store i32 [[TMP124]], i32* [[CONV37]], align 4
// CHECK17-NEXT:    [[TMP125:%.*]] = load i64, i64* [[N_CASTED36]], align 8
// CHECK17-NEXT:    [[TMP126:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK17-NEXT:    [[TMP127:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES41]] to i8*
// CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP127]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes.12 to i8*), i64 24, i1 false)
// CHECK17-NEXT:    [[TMP128:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP129:%.*]] = bitcast i8** [[TMP128]] to i64*
// CHECK17-NEXT:    store i64 [[TMP125]], i64* [[TMP129]], align 8
// CHECK17-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i64*
// CHECK17-NEXT:    store i64 [[TMP125]], i64* [[TMP131]], align 8
// CHECK17-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP132]], align 8
// CHECK17-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP134:%.*]] = bitcast i8** [[TMP133]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP134]], align 8
// CHECK17-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP136]], align 8
// CHECK17-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP137]], align 8
// CHECK17-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP139]], align 8
// CHECK17-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP141]], align 8
// CHECK17-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES41]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP126]], i64* [[TMP142]], align 8
// CHECK17-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP143]], align 8
// CHECK17-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES41]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP147:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP147]], i32* [[DOTCAPTURE_EXPR_43]], align 4
// CHECK17-NEXT:    [[TMP148:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_43]], align 4
// CHECK17-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[TMP148]], 0
// CHECK17-NEXT:    [[DIV46:%.*]] = sdiv i32 [[SUB45]], 1
// CHECK17-NEXT:    [[SUB47:%.*]] = sub nsw i32 [[DIV46]], 1
// CHECK17-NEXT:    store i32 [[SUB47]], i32* [[DOTCAPTURE_EXPR_44]], align 4
// CHECK17-NEXT:    [[TMP149:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_44]], align 4
// CHECK17-NEXT:    [[ADD48:%.*]] = add nsw i32 [[TMP149]], 1
// CHECK17-NEXT:    [[TMP150:%.*]] = zext i32 [[ADD48]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS49:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP151]], align 4
// CHECK17-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 1
// CHECK17-NEXT:    store i32 3, i32* [[TMP152]], align 4
// CHECK17-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP144]], i8*** [[TMP153]], align 8
// CHECK17-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP145]], i8*** [[TMP154]], align 8
// CHECK17-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP146]], i64** [[TMP155]], align 8
// CHECK17-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP156]], align 8
// CHECK17-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP157]], align 8
// CHECK17-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP158]], align 8
// CHECK17-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP150]], i64* [[TMP159]], align 8
// CHECK17-NEXT:    [[TMP160:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]])
// CHECK17-NEXT:    [[TMP161:%.*]] = icmp ne i32 [[TMP160]], 0
// CHECK17-NEXT:    br i1 [[TMP161]], label [[OMP_OFFLOAD_FAILED50:%.*]], label [[OMP_OFFLOAD_CONT51:%.*]]
// CHECK17:       omp_offload.failed50:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i64 [[TMP125]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT51]]
// CHECK17:       omp_offload.cont51:
// CHECK17-NEXT:    [[TMP162:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    [[CONV53:%.*]] = bitcast i64* [[M_CASTED52]] to i32*
// CHECK17-NEXT:    store i32 [[TMP162]], i32* [[CONV53]], align 4
// CHECK17-NEXT:    [[TMP163:%.*]] = load i64, i64* [[M_CASTED52]], align 8
// CHECK17-NEXT:    [[TMP164:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[CONV55:%.*]] = bitcast i64* [[N_CASTED54]] to i32*
// CHECK17-NEXT:    store i32 [[TMP164]], i32* [[CONV55]], align 4
// CHECK17-NEXT:    [[TMP165:%.*]] = load i64, i64* [[N_CASTED54]], align 8
// CHECK17-NEXT:    [[TMP166:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK17-NEXT:    [[TMP167:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES59]] to i8*
// CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP167]], i8* align 8 bitcast ([4 x i64]* @.offload_sizes.16 to i8*), i64 32, i1 false)
// CHECK17-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP169:%.*]] = bitcast i8** [[TMP168]] to i64*
// CHECK17-NEXT:    store i64 [[TMP163]], i64* [[TMP169]], align 8
// CHECK17-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP171:%.*]] = bitcast i8** [[TMP170]] to i64*
// CHECK17-NEXT:    store i64 [[TMP163]], i64* [[TMP171]], align 8
// CHECK17-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP172]], align 8
// CHECK17-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP174:%.*]] = bitcast i8** [[TMP173]] to i64*
// CHECK17-NEXT:    store i64 [[TMP165]], i64* [[TMP174]], align 8
// CHECK17-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP176:%.*]] = bitcast i8** [[TMP175]] to i64*
// CHECK17-NEXT:    store i64 [[TMP165]], i64* [[TMP176]], align 8
// CHECK17-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP177]], align 8
// CHECK17-NEXT:    [[TMP178:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP179:%.*]] = bitcast i8** [[TMP178]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP179]], align 8
// CHECK17-NEXT:    [[TMP180:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP181:%.*]] = bitcast i8** [[TMP180]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP181]], align 8
// CHECK17-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP182]], align 8
// CHECK17-NEXT:    [[TMP183:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP184:%.*]] = bitcast i8** [[TMP183]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP184]], align 8
// CHECK17-NEXT:    [[TMP185:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP186:%.*]] = bitcast i8** [[TMP185]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP186]], align 8
// CHECK17-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES59]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP166]], i64* [[TMP187]], align 8
// CHECK17-NEXT:    [[TMP188:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP188]], align 8
// CHECK17-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP190:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES59]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP192:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP192]], i32* [[DOTCAPTURE_EXPR_61]], align 4
// CHECK17-NEXT:    [[TMP193:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_61]], align 4
// CHECK17-NEXT:    [[SUB63:%.*]] = sub nsw i32 [[TMP193]], 0
// CHECK17-NEXT:    [[DIV64:%.*]] = sdiv i32 [[SUB63]], 1
// CHECK17-NEXT:    [[SUB65:%.*]] = sub nsw i32 [[DIV64]], 1
// CHECK17-NEXT:    store i32 [[SUB65]], i32* [[DOTCAPTURE_EXPR_62]], align 4
// CHECK17-NEXT:    [[TMP194:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_62]], align 4
// CHECK17-NEXT:    [[ADD66:%.*]] = add nsw i32 [[TMP194]], 1
// CHECK17-NEXT:    [[TMP195:%.*]] = zext i32 [[ADD66]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS67:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP196:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP196]], align 4
// CHECK17-NEXT:    [[TMP197:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 1
// CHECK17-NEXT:    store i32 4, i32* [[TMP197]], align 4
// CHECK17-NEXT:    [[TMP198:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP189]], i8*** [[TMP198]], align 8
// CHECK17-NEXT:    [[TMP199:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP190]], i8*** [[TMP199]], align 8
// CHECK17-NEXT:    [[TMP200:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP191]], i64** [[TMP200]], align 8
// CHECK17-NEXT:    [[TMP201:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP201]], align 8
// CHECK17-NEXT:    [[TMP202:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP202]], align 8
// CHECK17-NEXT:    [[TMP203:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP203]], align 8
// CHECK17-NEXT:    [[TMP204:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP195]], i64* [[TMP204]], align 8
// CHECK17-NEXT:    [[TMP205:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]])
// CHECK17-NEXT:    [[TMP206:%.*]] = icmp ne i32 [[TMP205]], 0
// CHECK17-NEXT:    br i1 [[TMP206]], label [[OMP_OFFLOAD_FAILED68:%.*]], label [[OMP_OFFLOAD_CONT69:%.*]]
// CHECK17:       omp_offload.failed68:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i64 [[TMP163]], i64 [[TMP165]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT69]]
// CHECK17:       omp_offload.cont69:
// CHECK17-NEXT:    [[TMP207:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK17-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP207]])
// CHECK17-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK17-NEXT:    [[TMP208:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP208]])
// CHECK17-NEXT:    [[TMP209:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK17-NEXT:    ret i32 [[TMP209]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
// CHECK17-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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 8
// 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
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// 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*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 8
// 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:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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 i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK17-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV3]], 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 8
// 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:    [[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_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
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 8
// 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:    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_l153
// CHECK17-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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 8
// 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
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// 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*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 8
// 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:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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 i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK17-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV3]], 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 8
// 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:    [[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_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
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 8
// 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:    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_l158
// CHECK17-SAME: (i64 noundef [[M:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// 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 i64, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// 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* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// 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
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// 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
// CHECK17-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK17-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP21]], i32* [[CONV7]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
// CHECK17-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK17:       cond.true12:
// CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    br label [[COND_END14:%.*]]
// CHECK17:       cond.false13:
// CHECK17-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END14]]
// CHECK17:       cond.end14:
// CHECK17-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
// CHECK17-NEXT:    store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// 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:    [[I6:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// 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 i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK17-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV5]], 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 8
// 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:    [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP7]], 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
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK17-NEXT:    br i1 [[CMP8]], 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
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 8
// 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:    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_l163
// CHECK17-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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*, i64, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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 8
// 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
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// 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*, i64, i64, i32*, i64, i32*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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 8
// 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:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// 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 i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK17-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV3]], 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 8
// 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 8
// 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 !15
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK17-NEXT:    br i1 [[CMP5]], 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 !15
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !15
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !15
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !15
// 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 !15
// CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK17-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![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:    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_l168
// CHECK17-SAME: (i64 noundef [[M:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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*, i64, i32*, i64)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// 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 i64, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// 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 8
// 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
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK17-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP20]], i32* [[CONV7]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// 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:    [[I6:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// 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 i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK17-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV5]], 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* [[CONV]], 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 8
// 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 8
// 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 !18
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    br i1 [[CMP7]], 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 !18
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// 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:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK17-NEXT:    store i32 [[ADD8]], 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_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK17-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[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 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED22:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[_TMP27:%.*]] = 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 8
// 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 8
// CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP4]], align 8
// 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK17-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 8
// CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 8
// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK17-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP11]], align 8
// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP12]], align 8
// CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP13]], align 8
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP14]], align 8
// CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK17-NEXT:    store i64 10, i64* [[TMP15]], align 8
// CHECK17-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK17-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK17-NEXT:    br i1 [[TMP17]], 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__l116([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 8
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 8
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK17-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK17-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 8
// CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 8
// CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK17-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.24, i32 0, i32 0), i64** [[TMP29]], align 8
// CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.25, i32 0, i32 0), i64** [[TMP30]], align 8
// CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP31]], align 8
// CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP32]], align 8
// CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK17-NEXT:    store i64 10, i64* [[TMP33]], align 8
// CHECK17-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
// CHECK17-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK17-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK17:       omp_offload.failed6:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK17:       omp_offload.cont7:
// CHECK17-NEXT:    [[TMP36:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP36]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP37:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK17-NEXT:    store i64 [[TMP37]], i64* [[TMP39]], align 8
// CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
// CHECK17-NEXT:    store i64 [[TMP37]], i64* [[TMP41]], align 8
// CHECK17-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP42]], align 8
// CHECK17-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 8
// CHECK17-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 8
// CHECK17-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP47]], align 8
// CHECK17-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK17-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP50]], align 4
// CHECK17-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
// CHECK17-NEXT:    store i32 2, i32* [[TMP51]], align 4
// CHECK17-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP48]], i8*** [[TMP52]], align 8
// CHECK17-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP49]], i8*** [[TMP53]], align 8
// CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
// CHECK17-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.28, i32 0, i32 0), i64** [[TMP54]], align 8
// CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.29, i32 0, i32 0), i64** [[TMP55]], align 8
// CHECK17-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP56]], align 8
// CHECK17-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP57]], align 8
// CHECK17-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
// CHECK17-NEXT:    store i64 10, i64* [[TMP58]], align 8
// CHECK17-NEXT:    [[TMP59:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
// CHECK17-NEXT:    [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
// CHECK17-NEXT:    br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
// CHECK17:       omp_offload.failed13:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i64 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
// CHECK17:       omp_offload.cont14:
// CHECK17-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP62]], align 8
// CHECK17-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP64]], align 8
// CHECK17-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS17]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP65]], align 8
// CHECK17-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK17-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK17-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 1
// CHECK17-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK17-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP66]], i8*** [[TMP70]], align 8
// CHECK17-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP67]], i8*** [[TMP71]], align 8
// CHECK17-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 4
// CHECK17-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.32, i32 0, i32 0), i64** [[TMP72]], align 8
// CHECK17-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.33, i32 0, i32 0), i64** [[TMP73]], align 8
// CHECK17-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP74]], align 8
// CHECK17-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP75]], align 8
// CHECK17-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 8
// CHECK17-NEXT:    store i64 10, i64* [[TMP76]], align 8
// CHECK17-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]])
// CHECK17-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK17-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
// CHECK17:       omp_offload.failed20:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
// CHECK17:       omp_offload.cont21:
// CHECK17-NEXT:    [[TMP79:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    [[CONV23:%.*]] = bitcast i64* [[M_CASTED22]] to i32*
// CHECK17-NEXT:    store i32 [[TMP79]], i32* [[CONV23]], align 4
// CHECK17-NEXT:    [[TMP80:%.*]] = load i64, i64* [[M_CASTED22]], align 8
// CHECK17-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
// CHECK17-NEXT:    store i64 [[TMP80]], i64* [[TMP82]], align 8
// CHECK17-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP84:%.*]] = bitcast i8** [[TMP83]] to i64*
// CHECK17-NEXT:    store i64 [[TMP80]], i64* [[TMP84]], align 8
// CHECK17-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP85]], align 8
// CHECK17-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP87]], align 8
// CHECK17-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP89]], align 8
// CHECK17-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP90]], align 8
// CHECK17-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK17-NEXT:    [[KERNEL_ARGS28:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP93]], align 4
// CHECK17-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 1
// CHECK17-NEXT:    store i32 2, i32* [[TMP94]], align 4
// CHECK17-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP91]], i8*** [[TMP95]], align 8
// CHECK17-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP92]], i8*** [[TMP96]], align 8
// CHECK17-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 4
// CHECK17-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.36, i32 0, i32 0), i64** [[TMP97]], align 8
// CHECK17-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.37, i32 0, i32 0), i64** [[TMP98]], align 8
// CHECK17-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP99]], align 8
// CHECK17-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP100]], align 8
// CHECK17-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 8
// CHECK17-NEXT:    store i64 10, i64* [[TMP101]], align 8
// CHECK17-NEXT:    [[TMP102:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]])
// CHECK17-NEXT:    [[TMP103:%.*]] = icmp ne i32 [[TMP102]], 0
// CHECK17-NEXT:    br i1 [[TMP103]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
// CHECK17:       omp_offload.failed29:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i64 [[TMP80]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT30]]
// CHECK17:       omp_offload.cont30:
// CHECK17-NEXT:    ret i32 0
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
// CHECK17-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// 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..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// 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 8
// 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
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// 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*, i64, i64, [10 x i32]*)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..19
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK17-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV1]], 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 8
// 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
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK17-NEXT:    br i1 [[CMP2]], 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
// 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
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
// CHECK17-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// 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..22 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// 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 8
// 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
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// 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*, i64, i64, [10 x i32]*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..23
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK17-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV1]], 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 8
// 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
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK17-NEXT:    br i1 [[CMP2]], 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
// 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
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
// CHECK17-SAME: (i64 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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]*, i64)* @.omp_outlined..26 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// 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 i64, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// 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 8
// 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
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..27 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..27
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK17-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV2]], 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* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// 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 i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
// 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 [ [[CONV4]], [[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:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP5]], 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
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK17-NEXT:    br i1 [[CMP6]], 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
// 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
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD8]], 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:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK17-NEXT:    store i32 [[ADD9]], 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
// CHECK17-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// 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..30 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// 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 8
// 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
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// 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*, i64, i64, [10 x i32]*)* @.omp_outlined..31 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..31
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK17-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV1]], 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 8
// 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 !21
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
// 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 !21
// 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 !21
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
// 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 !21
// CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK17-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
// CHECK17-SAME: (i64 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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]*, i64)* @.omp_outlined..34 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..34
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// 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 i64, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// 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 8
// 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
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK17-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// 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*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..35 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..35
// CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// 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 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK17-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK17-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[CONV2]], 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* [[CONV]], 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 8
// 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 !24
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
// 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 !24
// 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 !24
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !24
// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !24
// 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 !24
// CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK17-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK17-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK17-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@main
// CHECK19-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[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_CASTED18:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES22:%.*]] = alloca [4 x i64], align 4
// CHECK19-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED33:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS34:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS35:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS36:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES37:%.*]] = alloca [3 x i64], align 4
// CHECK19-NEXT:    [[_TMP38:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[M_CASTED48:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[N_CASTED49:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS50:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS51:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS52:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES53:%.*]] = alloca [4 x i64], align 4
// CHECK19-NEXT:    [[_TMP54:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_55:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_56:%.*]] = 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:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
// CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP8]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[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 i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP18]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP20]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK19-NEXT:    store i64 [[TMP5]], i64* [[TMP21]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP27]], 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:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK19-NEXT:    [[TMP29:%.*]] = zext i32 [[ADD]] to i64
// CHECK19-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP30]], align 4
// CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK19-NEXT:    store i32 3, i32* [[TMP31]], align 4
// CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP23]], i8*** [[TMP32]], align 4
// CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP24]], i8*** [[TMP33]], align 4
// CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK19-NEXT:    store i64* [[TMP25]], i64** [[TMP34]], align 4
// CHECK19-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
// CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP36]], align 4
// CHECK19-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP37]], align 4
// CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK19-NEXT:    store i64 [[TMP29]], i64* [[TMP38]], align 8
// CHECK19-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK19-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK19-NEXT:    br i1 [[TMP40]], 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_l148(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[TMP41:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP41]], i32* [[N_CASTED3]], align 4
// CHECK19-NEXT:    [[TMP42:%.*]] = load i32, i32* [[N_CASTED3]], align 4
// CHECK19-NEXT:    [[TMP43:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP44:%.*]] = sext i32 [[TMP43]] to i64
// CHECK19-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES7]] to i8*
// CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.4 to i8*), i32 24, i1 false)
// CHECK19-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
// CHECK19-NEXT:    store i32 [[TMP42]], i32* [[TMP47]], align 4
// CHECK19-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
// CHECK19-NEXT:    store i32 [[TMP42]], i32* [[TMP49]], align 4
// CHECK19-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP50]], align 4
// CHECK19-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP52]], align 4
// CHECK19-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP54:%.*]] = bitcast i8** [[TMP53]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP54]], align 4
// CHECK19-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP55]], align 4
// CHECK19-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP57]], align 4
// CHECK19-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP59]], align 4
// CHECK19-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
// CHECK19-NEXT:    store i64 [[TMP44]], i64* [[TMP60]], align 4
// CHECK19-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP61]], align 4
// CHECK19-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP65:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP65]], i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK19-NEXT:    [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK19-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP66]], 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:    [[TMP67:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK19-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP67]], 1
// CHECK19-NEXT:    [[TMP68:%.*]] = zext i32 [[ADD14]] to i64
// CHECK19-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK19-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK19-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK19-NEXT:    store i32 3, i32* [[TMP70]], align 4
// CHECK19-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP62]], i8*** [[TMP71]], align 4
// CHECK19-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP63]], i8*** [[TMP72]], align 4
// CHECK19-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK19-NEXT:    store i64* [[TMP64]], i64** [[TMP73]], align 4
// CHECK19-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP74]], align 4
// CHECK19-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP75]], align 4
// CHECK19-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP76]], align 4
// CHECK19-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK19-NEXT:    store i64 [[TMP68]], i64* [[TMP77]], align 8
// CHECK19-NEXT:    [[TMP78:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK19-NEXT:    [[TMP79:%.*]] = icmp ne i32 [[TMP78]], 0
// CHECK19-NEXT:    br i1 [[TMP79]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK19:       omp_offload.failed16:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i32 [[TMP42]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK19:       omp_offload.cont17:
// CHECK19-NEXT:    [[TMP80:%.*]] = load i32, i32* [[M]], align 4
// CHECK19-NEXT:    store i32 [[TMP80]], i32* [[M_CASTED]], align 4
// CHECK19-NEXT:    [[TMP81:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK19-NEXT:    [[TMP82:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP82]], i32* [[N_CASTED18]], align 4
// CHECK19-NEXT:    [[TMP83:%.*]] = load i32, i32* [[N_CASTED18]], align 4
// CHECK19-NEXT:    [[TMP84:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP85:%.*]] = sext i32 [[TMP84]] to i64
// CHECK19-NEXT:    [[TMP86:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES22]] to i8*
// CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP86]], i8* align 4 bitcast ([4 x i64]* @.offload_sizes.8 to i8*), i32 32, i1 false)
// CHECK19-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
// CHECK19-NEXT:    store i32 [[TMP81]], i32* [[TMP88]], align 4
// CHECK19-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP90:%.*]] = bitcast i8** [[TMP89]] to i32*
// CHECK19-NEXT:    store i32 [[TMP81]], i32* [[TMP90]], align 4
// CHECK19-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP91]], align 4
// CHECK19-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i32*
// CHECK19-NEXT:    store i32 [[TMP83]], i32* [[TMP93]], align 4
// CHECK19-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP95:%.*]] = bitcast i8** [[TMP94]] to i32*
// CHECK19-NEXT:    store i32 [[TMP83]], i32* [[TMP95]], align 4
// CHECK19-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP96]], align 4
// CHECK19-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP98]], align 4
// CHECK19-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP100]], align 4
// CHECK19-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP101]], align 4
// CHECK19-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP103]], align 4
// CHECK19-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP105]], align 4
// CHECK19-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 3
// CHECK19-NEXT:    store i64 [[TMP85]], i64* [[TMP106]], align 4
// CHECK19-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 3
// CHECK19-NEXT:    store i8* null, i8** [[TMP107]], align 4
// CHECK19-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP111:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP111]], i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK19-NEXT:    [[TMP112:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK19-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[TMP112]], 0
// CHECK19-NEXT:    [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
// CHECK19-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[DIV27]], 1
// CHECK19-NEXT:    store i32 [[SUB28]], i32* [[DOTCAPTURE_EXPR_25]], align 4
// CHECK19-NEXT:    [[TMP113:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
// CHECK19-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP113]], 1
// CHECK19-NEXT:    [[TMP114:%.*]] = zext i32 [[ADD29]] to i64
// CHECK19-NEXT:    [[KERNEL_ARGS30:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK19-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP115]], align 4
// CHECK19-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 1
// CHECK19-NEXT:    store i32 4, i32* [[TMP116]], align 4
// CHECK19-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP108]], i8*** [[TMP117]], align 4
// CHECK19-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP109]], i8*** [[TMP118]], align 4
// CHECK19-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 4
// CHECK19-NEXT:    store i64* [[TMP110]], i64** [[TMP119]], align 4
// CHECK19-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP120]], align 4
// CHECK19-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP121]], align 4
// CHECK19-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP122]], align 4
// CHECK19-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 8
// CHECK19-NEXT:    store i64 [[TMP114]], i64* [[TMP123]], align 8
// CHECK19-NEXT:    [[TMP124:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]])
// CHECK19-NEXT:    [[TMP125:%.*]] = icmp ne i32 [[TMP124]], 0
// CHECK19-NEXT:    br i1 [[TMP125]], label [[OMP_OFFLOAD_FAILED31:%.*]], label [[OMP_OFFLOAD_CONT32:%.*]]
// CHECK19:       omp_offload.failed31:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i32 [[TMP81]], i32 [[TMP83]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT32]]
// CHECK19:       omp_offload.cont32:
// CHECK19-NEXT:    [[TMP126:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP126]], i32* [[N_CASTED33]], align 4
// CHECK19-NEXT:    [[TMP127:%.*]] = load i32, i32* [[N_CASTED33]], align 4
// CHECK19-NEXT:    [[TMP128:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP129:%.*]] = sext i32 [[TMP128]] to i64
// CHECK19-NEXT:    [[TMP130:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES37]] to i8*
// CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP130]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.12 to i8*), i32 24, i1 false)
// CHECK19-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP132:%.*]] = bitcast i8** [[TMP131]] to i32*
// CHECK19-NEXT:    store i32 [[TMP127]], i32* [[TMP132]], align 4
// CHECK19-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP134:%.*]] = bitcast i8** [[TMP133]] to i32*
// CHECK19-NEXT:    store i32 [[TMP127]], i32* [[TMP134]], align 4
// CHECK19-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP135]], align 4
// CHECK19-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP137]], align 4
// CHECK19-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP139]], align 4
// CHECK19-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP140]], align 4
// CHECK19-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP142]], align 4
// CHECK19-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP144]], align 4
// CHECK19-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES37]], i32 0, i32 2
// CHECK19-NEXT:    store i64 [[TMP129]], i64* [[TMP145]], align 4
// CHECK19-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP146]], align 4
// CHECK19-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP149:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES37]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP150:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP150]], i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK19-NEXT:    [[TMP151:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK19-NEXT:    [[SUB41:%.*]] = sub nsw i32 [[TMP151]], 0
// CHECK19-NEXT:    [[DIV42:%.*]] = sdiv i32 [[SUB41]], 1
// CHECK19-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[DIV42]], 1
// CHECK19-NEXT:    store i32 [[SUB43]], i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK19-NEXT:    [[TMP152:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK19-NEXT:    [[ADD44:%.*]] = add nsw i32 [[TMP152]], 1
// CHECK19-NEXT:    [[TMP153:%.*]] = zext i32 [[ADD44]] to i64
// CHECK19-NEXT:    [[KERNEL_ARGS45:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK19-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP154]], align 4
// CHECK19-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 1
// CHECK19-NEXT:    store i32 3, i32* [[TMP155]], align 4
// CHECK19-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP147]], i8*** [[TMP156]], align 4
// CHECK19-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP148]], i8*** [[TMP157]], align 4
// CHECK19-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 4
// CHECK19-NEXT:    store i64* [[TMP149]], i64** [[TMP158]], align 4
// CHECK19-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP159]], align 4
// CHECK19-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP160]], align 4
// CHECK19-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP161]], align 4
// CHECK19-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 8
// CHECK19-NEXT:    store i64 [[TMP153]], i64* [[TMP162]], align 8
// CHECK19-NEXT:    [[TMP163:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]])
// CHECK19-NEXT:    [[TMP164:%.*]] = icmp ne i32 [[TMP163]], 0
// CHECK19-NEXT:    br i1 [[TMP164]], label [[OMP_OFFLOAD_FAILED46:%.*]], label [[OMP_OFFLOAD_CONT47:%.*]]
// CHECK19:       omp_offload.failed46:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i32 [[TMP127]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT47]]
// CHECK19:       omp_offload.cont47:
// CHECK19-NEXT:    [[TMP165:%.*]] = load i32, i32* [[M]], align 4
// CHECK19-NEXT:    store i32 [[TMP165]], i32* [[M_CASTED48]], align 4
// CHECK19-NEXT:    [[TMP166:%.*]] = load i32, i32* [[M_CASTED48]], align 4
// CHECK19-NEXT:    [[TMP167:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP167]], i32* [[N_CASTED49]], align 4
// CHECK19-NEXT:    [[TMP168:%.*]] = load i32, i32* [[N_CASTED49]], align 4
// CHECK19-NEXT:    [[TMP169:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK19-NEXT:    [[TMP170:%.*]] = sext i32 [[TMP169]] to i64
// CHECK19-NEXT:    [[TMP171:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES53]] to i8*
// CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP171]], i8* align 4 bitcast ([4 x i64]* @.offload_sizes.16 to i8*), i32 32, i1 false)
// CHECK19-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP173:%.*]] = bitcast i8** [[TMP172]] to i32*
// CHECK19-NEXT:    store i32 [[TMP166]], i32* [[TMP173]], align 4
// CHECK19-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP175:%.*]] = bitcast i8** [[TMP174]] to i32*
// CHECK19-NEXT:    store i32 [[TMP166]], i32* [[TMP175]], align 4
// CHECK19-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP176]], align 4
// CHECK19-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP178:%.*]] = bitcast i8** [[TMP177]] to i32*
// CHECK19-NEXT:    store i32 [[TMP168]], i32* [[TMP178]], align 4
// CHECK19-NEXT:    [[TMP179:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP180:%.*]] = bitcast i8** [[TMP179]] to i32*
// CHECK19-NEXT:    store i32 [[TMP168]], i32* [[TMP180]], align 4
// CHECK19-NEXT:    [[TMP181:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP181]], align 4
// CHECK19-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP183:%.*]] = bitcast i8** [[TMP182]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP183]], align 4
// CHECK19-NEXT:    [[TMP184:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP185:%.*]] = bitcast i8** [[TMP184]] to i32*
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP185]], align 4
// CHECK19-NEXT:    [[TMP186:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP186]], align 4
// CHECK19-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP188:%.*]] = bitcast i8** [[TMP187]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP188]], align 4
// CHECK19-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP190:%.*]] = bitcast i8** [[TMP189]] to i32**
// CHECK19-NEXT:    store i32* [[VLA]], i32** [[TMP190]], align 4
// CHECK19-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES53]], i32 0, i32 3
// CHECK19-NEXT:    store i64 [[TMP170]], i64* [[TMP191]], align 4
// CHECK19-NEXT:    [[TMP192:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 3
// CHECK19-NEXT:    store i8* null, i8** [[TMP192]], align 4
// CHECK19-NEXT:    [[TMP193:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP194:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP195:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES53]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP196:%.*]] = load i32, i32* [[N]], align 4
// CHECK19-NEXT:    store i32 [[TMP196]], i32* [[DOTCAPTURE_EXPR_55]], align 4
// CHECK19-NEXT:    [[TMP197:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_55]], align 4
// CHECK19-NEXT:    [[SUB57:%.*]] = sub nsw i32 [[TMP197]], 0
// CHECK19-NEXT:    [[DIV58:%.*]] = sdiv i32 [[SUB57]], 1
// CHECK19-NEXT:    [[SUB59:%.*]] = sub nsw i32 [[DIV58]], 1
// CHECK19-NEXT:    store i32 [[SUB59]], i32* [[DOTCAPTURE_EXPR_56]], align 4
// CHECK19-NEXT:    [[TMP198:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_56]], align 4
// CHECK19-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP198]], 1
// CHECK19-NEXT:    [[TMP199:%.*]] = zext i32 [[ADD60]] to i64
// CHECK19-NEXT:    [[KERNEL_ARGS61:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK19-NEXT:    [[TMP200:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP200]], align 4
// CHECK19-NEXT:    [[TMP201:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 1
// CHECK19-NEXT:    store i32 4, i32* [[TMP201]], align 4
// CHECK19-NEXT:    [[TMP202:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP193]], i8*** [[TMP202]], align 4
// CHECK19-NEXT:    [[TMP203:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP194]], i8*** [[TMP203]], align 4
// CHECK19-NEXT:    [[TMP204:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 4
// CHECK19-NEXT:    store i64* [[TMP195]], i64** [[TMP204]], align 4
// CHECK19-NEXT:    [[TMP205:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP205]], align 4
// CHECK19-NEXT:    [[TMP206:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP206]], align 4
// CHECK19-NEXT:    [[TMP207:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP207]], align 4
// CHECK19-NEXT:    [[TMP208:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 8
// CHECK19-NEXT:    store i64 [[TMP199]], i64* [[TMP208]], align 8
// CHECK19-NEXT:    [[TMP209:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]])
// CHECK19-NEXT:    [[TMP210:%.*]] = icmp ne i32 [[TMP209]], 0
// CHECK19-NEXT:    br i1 [[TMP210]], label [[OMP_OFFLOAD_FAILED62:%.*]], label [[OMP_OFFLOAD_CONT63:%.*]]
// CHECK19:       omp_offload.failed62:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i32 [[TMP166]], i32 [[TMP168]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT63]]
// CHECK19:       omp_offload.cont63:
// CHECK19-NEXT:    [[TMP211:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK19-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP211]])
// CHECK19-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK19-NEXT:    [[TMP212:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP212]])
// CHECK19-NEXT:    [[TMP213:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK19-NEXT:    ret i32 [[TMP213]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
// CHECK19-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    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_l153
// CHECK19-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    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_l158
// CHECK19-SAME: (i32 noundef [[M:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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..6 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[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
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// 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
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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..7 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK19-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK19-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// 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
// CHECK19-NEXT:    br label [[COND_END13:%.*]]
// CHECK19:       cond.false12:
// CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[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
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// 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
// CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    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_l163
// CHECK19-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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..10 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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..11 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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 35, 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 !16
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
// 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 !16
// 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 !16
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
// 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 !16
// CHECK19-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK19-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![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:    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_l168
// CHECK19-SAME: (i32 noundef [[M:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef 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..14 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[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
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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..15 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[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 35, 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 !19
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
// 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 !19
// 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 !19
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !19
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
// 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 !19
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![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:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK19-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[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_BASEPTRS8:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[M_CASTED22:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS23:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS24:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS25:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[_TMP26:%.*]] = 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK19-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK19-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP11]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP12]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP14]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK19-NEXT:    store i64 10, i64* [[TMP15]], align 8
// CHECK19-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK19-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK19-NEXT:    br i1 [[TMP17]], 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__l116([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK19-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK19-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 4
// CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 4
// CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK19-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.24, i32 0, i32 0), i64** [[TMP29]], align 4
// CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.25, i32 0, i32 0), i64** [[TMP30]], align 4
// CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP31]], align 4
// CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP32]], align 4
// CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK19-NEXT:    store i64 10, i64* [[TMP33]], align 8
// CHECK19-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
// CHECK19-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK19-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK19:       omp_offload.failed6:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK19:       omp_offload.cont7:
// CHECK19-NEXT:    [[TMP36:%.*]] = load i32, i32* [[M]], align 4
// CHECK19-NEXT:    store i32 [[TMP36]], i32* [[M_CASTED]], align 4
// CHECK19-NEXT:    [[TMP37:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i32*
// CHECK19-NEXT:    store i32 [[TMP37]], i32* [[TMP39]], align 4
// CHECK19-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i32*
// CHECK19-NEXT:    store i32 [[TMP37]], i32* [[TMP41]], align 4
// CHECK19-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP42]], align 4
// CHECK19-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 4
// CHECK19-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 4
// CHECK19-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP47]], align 4
// CHECK19-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK19-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK19-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP50]], align 4
// CHECK19-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
// CHECK19-NEXT:    store i32 2, i32* [[TMP51]], align 4
// CHECK19-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP48]], i8*** [[TMP52]], align 4
// CHECK19-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP49]], i8*** [[TMP53]], align 4
// CHECK19-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
// CHECK19-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.28, i32 0, i32 0), i64** [[TMP54]], align 4
// CHECK19-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.29, i32 0, i32 0), i64** [[TMP55]], align 4
// CHECK19-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP56]], align 4
// CHECK19-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP57]], align 4
// CHECK19-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
// CHECK19-NEXT:    store i64 10, i64* [[TMP58]], align 8
// CHECK19-NEXT:    [[TMP59:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
// CHECK19-NEXT:    [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
// CHECK19-NEXT:    br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
// CHECK19:       omp_offload.failed13:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i32 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
// CHECK19:       omp_offload.cont14:
// CHECK19-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP62]], align 4
// CHECK19-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP64]], align 4
// CHECK19-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS17]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP65]], align 4
// CHECK19-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK19-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK19-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK19-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 1
// CHECK19-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK19-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP66]], i8*** [[TMP70]], align 4
// CHECK19-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP67]], i8*** [[TMP71]], align 4
// CHECK19-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 4
// CHECK19-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.32, i32 0, i32 0), i64** [[TMP72]], align 4
// CHECK19-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.33, i32 0, i32 0), i64** [[TMP73]], align 4
// CHECK19-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP74]], align 4
// CHECK19-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP75]], align 4
// CHECK19-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 8
// CHECK19-NEXT:    store i64 10, i64* [[TMP76]], align 8
// CHECK19-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]])
// CHECK19-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK19-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
// CHECK19:       omp_offload.failed20:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
// CHECK19:       omp_offload.cont21:
// CHECK19-NEXT:    [[TMP79:%.*]] = load i32, i32* [[M]], align 4
// CHECK19-NEXT:    store i32 [[TMP79]], i32* [[M_CASTED22]], align 4
// CHECK19-NEXT:    [[TMP80:%.*]] = load i32, i32* [[M_CASTED22]], align 4
// CHECK19-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
// CHECK19-NEXT:    store i32 [[TMP80]], i32* [[TMP82]], align 4
// CHECK19-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP84:%.*]] = bitcast i8** [[TMP83]] to i32*
// CHECK19-NEXT:    store i32 [[TMP80]], i32* [[TMP84]], align 4
// CHECK19-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP85]], align 4
// CHECK19-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP87]], align 4
// CHECK19-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP89]], align 4
// CHECK19-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP90]], align 4
// CHECK19-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
// CHECK19-NEXT:    [[KERNEL_ARGS27:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK19-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 0
// CHECK19-NEXT:    store i32 1, i32* [[TMP93]], align 4
// CHECK19-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 1
// CHECK19-NEXT:    store i32 2, i32* [[TMP94]], align 4
// CHECK19-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 2
// CHECK19-NEXT:    store i8** [[TMP91]], i8*** [[TMP95]], align 4
// CHECK19-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 3
// CHECK19-NEXT:    store i8** [[TMP92]], i8*** [[TMP96]], align 4
// CHECK19-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 4
// CHECK19-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.36, i32 0, i32 0), i64** [[TMP97]], align 4
// CHECK19-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 5
// CHECK19-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.37, i32 0, i32 0), i64** [[TMP98]], align 4
// CHECK19-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 6
// CHECK19-NEXT:    store i8** null, i8*** [[TMP99]], align 4
// CHECK19-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 7
// CHECK19-NEXT:    store i8** null, i8*** [[TMP100]], align 4
// CHECK19-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 8
// CHECK19-NEXT:    store i64 10, i64* [[TMP101]], align 8
// CHECK19-NEXT:    [[TMP102:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]])
// CHECK19-NEXT:    [[TMP103:%.*]] = icmp ne i32 [[TMP102]], 0
// CHECK19-NEXT:    br i1 [[TMP103]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK19:       omp_offload.failed28:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i32 [[TMP80]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK19:       omp_offload.cont29:
// CHECK19-NEXT:    ret i32 0
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
// CHECK19-SAME: ([10 x i32]* noundef 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..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef 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
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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..19 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..19
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef 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
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
// CHECK19-SAME: ([10 x i32]* noundef 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..22 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef 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
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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..23 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..23
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef 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
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
// CHECK19-SAME: (i32 noundef [[M:%.*]], [10 x i32]* noundef 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..26 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[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
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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..27 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..27
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[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
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// 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
// 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
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
// 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
// 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
// CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK19-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
// CHECK19-SAME: ([10 x i32]* noundef 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..30 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef 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
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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..31 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..31
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef 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 35, 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 !22
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
// 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 !22
// 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 !22
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !22
// 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 !22
// 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 !22
// CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK19-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
// CHECK19-SAME: (i32 noundef [[M:%.*]], [10 x i32]* noundef 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..34 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..34
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[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
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// 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
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// 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..35 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
// 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..35
// CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[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 35, 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 !25
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
// 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 !25
// 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 !25
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !25
// 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 !25
// 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 !25
// CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK19-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK19-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK19-NEXT:    ret void
//