// 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 -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -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=CHECK2
// 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=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=CHECK2
// 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=CHECK5
// 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=CHECK6

// 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=CHECK5
// 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=CHECK6

// 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 --check-prefix=CHECK9
// 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 --check-prefix=CHECK9
// 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 --check-prefix=CHECK11
// 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 --check-prefix=CHECK11
#ifdef CK1

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

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

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

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


    return a[0];
  }
};

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

}
#endif // CK1

// Test host codegen.
// RUN: %clang_cc1 -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=CHECK13

// 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=CHECK14

// 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=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=CHECK14


// 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=CHECK17
// 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=CHECK17

// 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=CHECK19
// 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=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 --check-prefix=CHECK21
// 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 --check-prefix=CHECK21
// 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 --check-prefix=CHECK23
// 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 --check-prefix=CHECK23
#ifdef CK2

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

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


#endif // CK2
#endif // #ifndef HEADER
// CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK1-NEXT:    [[CALL:%.*]] = call 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_l36.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_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[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_l41.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_l41(%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_l46.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_l46(%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_l52.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_l52(%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_l58.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_l58(%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_l36
// 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, !llvm.access.group !8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !8
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias 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, !llvm.access.group !12
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK1-SAME: (%struct.SS* 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, !llvm.access.group !17
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !17
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK1-SAME: (i32* noalias 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, !llvm.access.group !20
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK1-SAME: (%struct.SS* 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, !llvm.access.group !23
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !23
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK1-SAME: (i32* noalias 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, !llvm.access.group !26
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK1:       omp.dispatch.inc:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK1:       omp.dispatch.end:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK1-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK1-SAME: (%struct.SS* 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, !llvm.access.group !29
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !29
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK1-SAME: (i32* noalias 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 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK1:       omp.dispatch.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK1:       omp.dispatch.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK1:       omp.dispatch.inc:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK1:       omp.dispatch.end:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK1-SAME: (%struct.SS* 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, !llvm.access.group !35
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !35
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK1-SAME: (i32* noalias 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 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK1:       omp.dispatch.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK1:       omp.dispatch.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK1:       omp.dispatch.inc:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK1:       omp.dispatch.end:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK2-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
// CHECK2-NEXT:    ret i32 [[CALL]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK2-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[_TMP30:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK2-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP14]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK2-NEXT:    store i64 123, i64* [[TMP15]], align 8
// CHECK2-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_l36.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK2-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK2-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK2-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
// CHECK2-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP31]], align 8
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP32]], align 8
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
// CHECK2-NEXT:    store i64 123, i64* [[TMP33]], align 8
// CHECK2-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_l41.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
// CHECK2-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK2-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK2:       omp_offload.failed8:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK2:       omp_offload.cont9:
// CHECK2-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
// CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 8
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP40]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK2-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK2-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP43]], align 4
// CHECK2-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK2-NEXT:    store i32 1, i32* [[TMP44]], align 4
// CHECK2-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP41]], i8*** [[TMP45]], align 8
// CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP42]], i8*** [[TMP46]], align 8
// CHECK2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 8
// CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 8
// CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP49]], align 8
// CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP50]], align 8
// CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK2-NEXT:    store i64 123, i64* [[TMP51]], align 8
// CHECK2-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_l46.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK2-NEXT:    [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
// CHECK2-NEXT:    br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK2:       omp_offload.failed16:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK2:       omp_offload.cont17:
// CHECK2-NEXT:    [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP55]], align 8
// CHECK2-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A18]], [123 x i32]** [[TMP57]], align 8
// CHECK2-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP58]], align 8
// CHECK2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK2-NEXT:    [[KERNEL_ARGS23:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK2-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP61]], align 4
// CHECK2-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 1
// CHECK2-NEXT:    store i32 1, i32* [[TMP62]], align 4
// CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP59]], i8*** [[TMP63]], align 8
// CHECK2-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP60]], i8*** [[TMP64]], align 8
// CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 4
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP65]], align 8
// CHECK2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP66]], align 8
// CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP67]], align 8
// CHECK2-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP68]], align 8
// CHECK2-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 8
// CHECK2-NEXT:    store i64 123, i64* [[TMP69]], align 8
// CHECK2-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_l52.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]])
// CHECK2-NEXT:    [[TMP71:%.*]] = icmp ne i32 [[TMP70]], 0
// CHECK2-NEXT:    br i1 [[TMP71]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK2:       omp_offload.failed24:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK2:       omp_offload.cont25:
// CHECK2-NEXT:    [[A26:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to %struct.SS**
// CHECK2-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP73]], align 8
// CHECK2-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [123 x i32]**
// CHECK2-NEXT:    store [123 x i32]* [[A26]], [123 x i32]** [[TMP75]], align 8
// CHECK2-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP76]], align 8
// CHECK2-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK2-NEXT:    [[KERNEL_ARGS31:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK2-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP79]], align 4
// CHECK2-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 1
// CHECK2-NEXT:    store i32 1, i32* [[TMP80]], align 4
// CHECK2-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP77]], i8*** [[TMP81]], align 8
// CHECK2-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP78]], i8*** [[TMP82]], align 8
// CHECK2-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 4
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP83]], align 8
// CHECK2-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP84]], align 8
// CHECK2-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP85]], align 8
// CHECK2-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP86]], align 8
// CHECK2-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 8
// CHECK2-NEXT:    store i64 123, i64* [[TMP87]], align 8
// CHECK2-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_l58.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]])
// CHECK2-NEXT:    [[TMP89:%.*]] = icmp ne i32 [[TMP88]], 0
// CHECK2-NEXT:    br i1 [[TMP89]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK2:       omp_offload.failed32:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK2:       omp_offload.cont33:
// CHECK2-NEXT:    [[A34:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A34]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP90:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    ret i32 [[TMP90]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK2-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !8
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK2-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !17
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK2-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !23
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK2:       omp.dispatch.cond:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK2:       omp.dispatch.body:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK2:       omp.dispatch.inc:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK2:       omp.dispatch.end:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK2-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK2-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !29
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK2:       omp.dispatch.cond:
// CHECK2-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK2:       omp.dispatch.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK2:       omp.dispatch.inc:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK2:       omp.dispatch.end:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK2-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !35
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK2:       omp.dispatch.cond:
// CHECK2-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK2:       omp.dispatch.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK2:       omp.dispatch.inc:
// CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK2:       omp.dispatch.end:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2:       .omp.final.then:
// CHECK2-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK2:       .omp.final.done:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK2-NEXT:    ret void
//
//
// 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 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 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 4
// CHECK5-NEXT:    [[_TMP30:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK5-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK5-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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_l36.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_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK5:       omp_offload.cont:
// CHECK5-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[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 4
// 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 4
// CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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_l41.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_l41(%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 4
// 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 4
// CHECK5-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP40]], align 4
// CHECK5-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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_l46.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_l46(%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 4
// 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 4
// CHECK5-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP58]], align 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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_l52.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_l52(%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 4
// 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 4
// CHECK5-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP76]], align 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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_l58.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_l58(%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]], i32 0, i32 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_l36
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !9
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK5-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !18
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK5-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !24
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK5:       omp.dispatch.cond:
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK5:       omp.dispatch.body:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK5:       omp.dispatch.inc:
// CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK5-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK5-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK5:       omp.dispatch.end:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK5-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK5-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !30
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK5-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK5:       omp.dispatch.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK5:       omp.dispatch.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK5:       omp.dispatch.inc:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK5:       omp.dispatch.end:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK5-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK5-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !36
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK5-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK5:       omp.dispatch.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK5:       omp.dispatch.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK5:       omp.dispatch.inc:
// CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK5:       omp.dispatch.end:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK5-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK5-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK5-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK6-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK6-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
// CHECK6-NEXT:    ret i32 [[CALL]]
//
//
// CHECK6-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK6-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[_TMP22:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 4
// CHECK6-NEXT:    [[_TMP30:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK6-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK6-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK6-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK6-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK6-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK6-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 4
// CHECK6-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK6-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 4
// CHECK6-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
// CHECK6-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK6-NEXT:    store i8** null, i8*** [[TMP13]], align 4
// CHECK6-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK6-NEXT:    store i8** null, i8*** [[TMP14]], align 4
// CHECK6-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK6-NEXT:    store i64 123, i64* [[TMP15]], align 8
// CHECK6-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_l36.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK6-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK6-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK6:       omp_offload.failed:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK6:       omp_offload.cont:
// CHECK6-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
// CHECK6-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 4
// CHECK6-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK6-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK6-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK6-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
// CHECK6-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK6-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
// CHECK6-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK6-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
// CHECK6-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 4
// CHECK6-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
// CHECK6-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 4
// CHECK6-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 4
// CHECK6-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 4
// CHECK6-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
// CHECK6-NEXT:    store i8** null, i8*** [[TMP31]], align 4
// CHECK6-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
// CHECK6-NEXT:    store i8** null, i8*** [[TMP32]], align 4
// CHECK6-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
// CHECK6-NEXT:    store i64 123, i64* [[TMP33]], align 8
// CHECK6-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_l41.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
// CHECK6-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK6-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK6:       omp_offload.failed8:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK6:       omp_offload.cont9:
// CHECK6-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
// CHECK6-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 4
// CHECK6-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP40]], align 4
// CHECK6-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK6-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK6-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK6-NEXT:    store i32 1, i32* [[TMP43]], align 4
// CHECK6-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK6-NEXT:    store i32 1, i32* [[TMP44]], align 4
// CHECK6-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK6-NEXT:    store i8** [[TMP41]], i8*** [[TMP45]], align 4
// CHECK6-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK6-NEXT:    store i8** [[TMP42]], i8*** [[TMP46]], align 4
// CHECK6-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 4
// CHECK6-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 4
// CHECK6-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK6-NEXT:    store i8** null, i8*** [[TMP49]], align 4
// CHECK6-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK6-NEXT:    store i8** null, i8*** [[TMP50]], align 4
// CHECK6-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK6-NEXT:    store i64 123, i64* [[TMP51]], align 8
// CHECK6-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_l46.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK6-NEXT:    [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
// CHECK6-NEXT:    br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK6:       omp_offload.failed16:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK6:       omp_offload.cont17:
// CHECK6-NEXT:    [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP55]], align 4
// CHECK6-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A18]], [123 x i32]** [[TMP57]], align 4
// CHECK6-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP58]], align 4
// CHECK6-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK6-NEXT:    [[KERNEL_ARGS23:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK6-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 0
// CHECK6-NEXT:    store i32 1, i32* [[TMP61]], align 4
// CHECK6-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 1
// CHECK6-NEXT:    store i32 1, i32* [[TMP62]], align 4
// CHECK6-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 2
// CHECK6-NEXT:    store i8** [[TMP59]], i8*** [[TMP63]], align 4
// CHECK6-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 3
// CHECK6-NEXT:    store i8** [[TMP60]], i8*** [[TMP64]], align 4
// CHECK6-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 4
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP65]], align 4
// CHECK6-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 5
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP66]], align 4
// CHECK6-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 6
// CHECK6-NEXT:    store i8** null, i8*** [[TMP67]], align 4
// CHECK6-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 7
// CHECK6-NEXT:    store i8** null, i8*** [[TMP68]], align 4
// CHECK6-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]], i32 0, i32 8
// CHECK6-NEXT:    store i64 123, i64* [[TMP69]], align 8
// CHECK6-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_l52.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS23]])
// CHECK6-NEXT:    [[TMP71:%.*]] = icmp ne i32 [[TMP70]], 0
// CHECK6-NEXT:    br i1 [[TMP71]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
// CHECK6:       omp_offload.failed24:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT25]]
// CHECK6:       omp_offload.cont25:
// CHECK6-NEXT:    [[A26:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to %struct.SS**
// CHECK6-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP73]], align 4
// CHECK6-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [123 x i32]**
// CHECK6-NEXT:    store [123 x i32]* [[A26]], [123 x i32]** [[TMP75]], align 4
// CHECK6-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i32 0, i32 0
// CHECK6-NEXT:    store i8* null, i8** [[TMP76]], align 4
// CHECK6-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
// CHECK6-NEXT:    [[KERNEL_ARGS31:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK6-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 0
// CHECK6-NEXT:    store i32 1, i32* [[TMP79]], align 4
// CHECK6-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 1
// CHECK6-NEXT:    store i32 1, i32* [[TMP80]], align 4
// CHECK6-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 2
// CHECK6-NEXT:    store i8** [[TMP77]], i8*** [[TMP81]], align 4
// CHECK6-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 3
// CHECK6-NEXT:    store i8** [[TMP78]], i8*** [[TMP82]], align 4
// CHECK6-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 4
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP83]], align 4
// CHECK6-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 5
// CHECK6-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP84]], align 4
// CHECK6-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 6
// CHECK6-NEXT:    store i8** null, i8*** [[TMP85]], align 4
// CHECK6-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 7
// CHECK6-NEXT:    store i8** null, i8*** [[TMP86]], align 4
// CHECK6-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]], i32 0, i32 8
// CHECK6-NEXT:    store i64 123, i64* [[TMP87]], align 8
// CHECK6-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_l58.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS31]])
// CHECK6-NEXT:    [[TMP89:%.*]] = icmp ne i32 [[TMP88]], 0
// CHECK6-NEXT:    br i1 [[TMP89]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
// CHECK6:       omp_offload.failed32:
// CHECK6-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58(%struct.SS* [[THIS1]]) #[[ATTR2]]
// CHECK6-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
// CHECK6:       omp_offload.cont33:
// CHECK6-NEXT:    [[A34:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A34]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP90:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK6-NEXT:    ret i32 [[TMP90]]
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
// CHECK6-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !9
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK6-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK6-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l41
// CHECK6-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !18
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK6-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK6-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l46
// CHECK6-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !24
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK6:       omp.dispatch.cond:
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK6:       omp.dispatch.body:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK6-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK6-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK6:       omp.dispatch.inc:
// CHECK6-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK6-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK6-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK6:       omp.dispatch.end:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK6-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK6-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l52
// CHECK6-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !30
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK6-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK6:       omp.dispatch.cond:
// CHECK6-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK6-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK6-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK6:       omp.dispatch.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK6-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK6:       omp.dispatch.inc:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK6:       omp.dispatch.end:
// CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK6-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l58
// CHECK6-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
// CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK6:       cond.true:
// CHECK6-NEXT:    br label [[COND_END:%.*]]
// CHECK6:       cond.false:
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    br label [[COND_END]]
// CHECK6:       cond.end:
// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !36
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK6:       omp.loop.exit:
// CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK6-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK6-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK6-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK6:       omp.dispatch.cond:
// CHECK6-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK6-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK6-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK6:       omp.dispatch.body:
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK6-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK6:       omp.inner.for.cond:
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK6:       omp.inner.for.body:
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK6:       omp.body.continue:
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK6:       omp.inner.for.inc:
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK6-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
// CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK6:       omp.inner.for.end:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK6:       omp.dispatch.inc:
// CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK6:       omp.dispatch.end:
// CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK6-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK6-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK6:       .omp.final.then:
// CHECK6-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK6:       .omp.final.done:
// CHECK6-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK6-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK6-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
// CHECK9-NEXT:    ret i32 [[CALL]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK9-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I7:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP20:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV23:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I24:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP37:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB38:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB39:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV40:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I41:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP54:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB55:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB56:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IV57:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I58:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK9-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB5]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
// CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
// CHECK9:       omp.inner.for.cond8:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
// CHECK9:       omp.inner.for.body10:
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK9-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK9-NEXT:    store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[IDXPROM14:%.*]] = sext i32 [[TMP10]] to i64
// CHECK9-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i64 0, i64 [[IDXPROM14]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX15]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE16:%.*]]
// CHECK9:       omp.body.continue16:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC17:%.*]]
// CHECK9:       omp.inner.for.inc17:
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK9:       omp.inner.for.end19:
// CHECK9-NEXT:    store i32 123, i32* [[I7]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB21]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB22]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB21]], align 4
// CHECK9-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV23]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND25:%.*]]
// CHECK9:       omp.inner.for.cond25:
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB22]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[CMP26:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK9-NEXT:    br i1 [[CMP26]], label [[OMP_INNER_FOR_BODY27:%.*]], label [[OMP_INNER_FOR_END36:%.*]]
// CHECK9:       omp.inner.for.body27:
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[MUL28:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK9-NEXT:    [[ADD29:%.*]] = add nsw i32 0, [[MUL28]]
// CHECK9-NEXT:    store i32 [[ADD29]], i32* [[I24]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I24]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[IDXPROM31:%.*]] = sext i32 [[TMP16]] to i64
// CHECK9-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 [[IDXPROM31]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX32]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE33:%.*]]
// CHECK9:       omp.body.continue33:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC34:%.*]]
// CHECK9:       omp.inner.for.inc34:
// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    [[ADD35:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK9-NEXT:    store i32 [[ADD35]], i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND25]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK9:       omp.inner.for.end36:
// CHECK9-NEXT:    store i32 123, i32* [[I24]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB38]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB39]], align 4
// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB38]], align 4
// CHECK9-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV40]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND42:%.*]]
// CHECK9:       omp.inner.for.cond42:
// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB39]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[CMP43:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK9-NEXT:    br i1 [[CMP43]], label [[OMP_INNER_FOR_BODY44:%.*]], label [[OMP_INNER_FOR_END53:%.*]]
// CHECK9:       omp.inner.for.body44:
// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[MUL45:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK9-NEXT:    [[ADD46:%.*]] = add nsw i32 0, [[MUL45]]
// CHECK9-NEXT:    store i32 [[ADD46]], i32* [[I41]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[A47:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I41]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[IDXPROM48:%.*]] = sext i32 [[TMP22]] to i64
// CHECK9-NEXT:    [[ARRAYIDX49:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A47]], i64 0, i64 [[IDXPROM48]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX49]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE50:%.*]]
// CHECK9:       omp.body.continue50:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC51:%.*]]
// CHECK9:       omp.inner.for.inc51:
// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    [[ADD52:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK9-NEXT:    store i32 [[ADD52]], i32* [[DOTOMP_IV40]], align 4, !llvm.access.group !12
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND42]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK9:       omp.inner.for.end53:
// CHECK9-NEXT:    store i32 123, i32* [[I41]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB55]], align 4
// CHECK9-NEXT:    store i32 122, i32* [[DOTOMP_UB56]], align 4
// CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB55]], align 4
// CHECK9-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV57]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND59:%.*]]
// CHECK9:       omp.inner.for.cond59:
// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB56]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[CMP60:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK9-NEXT:    br i1 [[CMP60]], label [[OMP_INNER_FOR_BODY61:%.*]], label [[OMP_INNER_FOR_END70:%.*]]
// CHECK9:       omp.inner.for.body61:
// CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[MUL62:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK9-NEXT:    [[ADD63:%.*]] = add nsw i32 0, [[MUL62]]
// CHECK9-NEXT:    store i32 [[ADD63]], i32* [[I58]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[A64:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[I58]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[IDXPROM65:%.*]] = sext i32 [[TMP28]] to i64
// CHECK9-NEXT:    [[ARRAYIDX66:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A64]], i64 0, i64 [[IDXPROM65]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX66]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE67:%.*]]
// CHECK9:       omp.body.continue67:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC68:%.*]]
// CHECK9:       omp.inner.for.inc68:
// CHECK9-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    [[ADD69:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK9-NEXT:    store i32 [[ADD69]], i32* [[DOTOMP_IV57]], align 4, !llvm.access.group !15
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND59]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK9:       omp.inner.for.end70:
// CHECK9-NEXT:    store i32 123, i32* [[I58]], align 4
// CHECK9-NEXT:    [[A71:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[ARRAYIDX72:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A71]], i64 0, i64 0
// CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[ARRAYIDX72]], align 4
// CHECK9-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
// CHECK11-NEXT:    ret i32 [[CALL]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK11-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I7:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP19:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB20:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB21:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV22:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I23:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP35:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB36:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB37:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV38:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I39:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB52:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB53:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV54:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I55:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP4]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK11-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    store i32 123, i32* [[I]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB5]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
// CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
// CHECK11:       omp.inner.for.cond8:
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK11-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
// CHECK11:       omp.inner.for.body10:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK11-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK11-NEXT:    store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i32 0, i32 [[TMP10]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX14]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
// CHECK11:       omp.body.continue15:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
// CHECK11:       omp.inner.for.inc16:
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK11-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK11:       omp.inner.for.end18:
// CHECK11-NEXT:    store i32 123, i32* [[I7]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB20]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB21]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB20]], align 4
// CHECK11-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV22]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND24:%.*]]
// CHECK11:       omp.inner.for.cond24:
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB21]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[CMP25:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK11-NEXT:    br i1 [[CMP25]], label [[OMP_INNER_FOR_BODY26:%.*]], label [[OMP_INNER_FOR_END34:%.*]]
// CHECK11:       omp.inner.for.body26:
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[MUL27:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK11-NEXT:    [[ADD28:%.*]] = add nsw i32 0, [[MUL27]]
// CHECK11-NEXT:    store i32 [[ADD28]], i32* [[I23]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[A29:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I23]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[ARRAYIDX30:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A29]], i32 0, i32 [[TMP16]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX30]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE31:%.*]]
// CHECK11:       omp.body.continue31:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC32:%.*]]
// CHECK11:       omp.inner.for.inc32:
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    [[ADD33:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK11-NEXT:    store i32 [[ADD33]], i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND24]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK11:       omp.inner.for.end34:
// CHECK11-NEXT:    store i32 123, i32* [[I23]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB36]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB37]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB36]], align 4
// CHECK11-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV38]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND40:%.*]]
// CHECK11:       omp.inner.for.cond40:
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB37]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[CMP41:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK11-NEXT:    br i1 [[CMP41]], label [[OMP_INNER_FOR_BODY42:%.*]], label [[OMP_INNER_FOR_END50:%.*]]
// CHECK11:       omp.inner.for.body42:
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[MUL43:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK11-NEXT:    [[ADD44:%.*]] = add nsw i32 0, [[MUL43]]
// CHECK11-NEXT:    store i32 [[ADD44]], i32* [[I39]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[A45:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I39]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[ARRAYIDX46:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A45]], i32 0, i32 [[TMP22]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX46]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE47:%.*]]
// CHECK11:       omp.body.continue47:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC48:%.*]]
// CHECK11:       omp.inner.for.inc48:
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    [[ADD49:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK11-NEXT:    store i32 [[ADD49]], i32* [[DOTOMP_IV38]], align 4, !llvm.access.group !13
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND40]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK11:       omp.inner.for.end50:
// CHECK11-NEXT:    store i32 123, i32* [[I39]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB52]], align 4
// CHECK11-NEXT:    store i32 122, i32* [[DOTOMP_UB53]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB52]], align 4
// CHECK11-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV54]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND56:%.*]]
// CHECK11:       omp.inner.for.cond56:
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB53]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[CMP57:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK11-NEXT:    br i1 [[CMP57]], label [[OMP_INNER_FOR_BODY58:%.*]], label [[OMP_INNER_FOR_END66:%.*]]
// CHECK11:       omp.inner.for.body58:
// CHECK11-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[MUL59:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK11-NEXT:    [[ADD60:%.*]] = add nsw i32 0, [[MUL59]]
// CHECK11-NEXT:    store i32 [[ADD60]], i32* [[I55]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[A61:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP28:%.*]] = load i32, i32* [[I55]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[ARRAYIDX62:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A61]], i32 0, i32 [[TMP28]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX62]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE63:%.*]]
// CHECK11:       omp.body.continue63:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC64:%.*]]
// CHECK11:       omp.inner.for.inc64:
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    [[ADD65:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK11-NEXT:    store i32 [[ADD65]], i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !16
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND56]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK11:       omp.inner.for.end66:
// CHECK11-NEXT:    store i32 123, i32* [[I55]], align 4
// CHECK11-NEXT:    [[A67:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[ARRAYIDX68:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A67]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[ARRAYIDX68]], align 4
// CHECK11-NEXT:    ret i32 [[TMP30]]
//
//
// 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_l154.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_l154(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_l159.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_l159(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_l164.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_l164(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_l169.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_l169(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_l174.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_l174(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_l154
// 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, !llvm.access.group !13
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !13
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK13-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK13-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK13-SAME: (i32* noalias 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, !llvm.access.group !17
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK13-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK13-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK13-SAME: (i64 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, !llvm.access.group !22
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !22
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK13-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK13-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK13-SAME: (i32* noalias 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, !llvm.access.group !25
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK13-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK13-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK13-SAME: (i64 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, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP21]], i32* [[CONV7]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !28
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]]), !llvm.access.group !28
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
// CHECK13-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK13:       cond.true12:
// CHECK13-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    br label [[COND_END14:%.*]]
// CHECK13:       cond.false13:
// CHECK13-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    br label [[COND_END14]]
// CHECK13:       cond.end14:
// CHECK13-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
// CHECK13-NEXT:    store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
// CHECK13-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK13-NEXT:    br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK13-NEXT:    [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
// CHECK13-NEXT:    [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD18]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..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, !llvm.access.group !31
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK13-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK13-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK13-NEXT:    [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
// CHECK13-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
// CHECK13-NEXT:    [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
// CHECK13-NEXT:    store i32 [[ADD13]], i32* [[I6]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK13-SAME: (i64 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, !llvm.access.group !34
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !34
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK13-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK13-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..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 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK13-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK13-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK13-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK13-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK13-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK13-SAME: (i64 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, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP20]], i32* [[CONV7]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !40
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]]), !llvm.access.group !40
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
// CHECK13-NEXT:    br i1 [[TMP27]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK13-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..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 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK13-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK13-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK13-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK13-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK13-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
// CHECK13-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
// CHECK13-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK13-NEXT:    store i32 [[ADD12]], i32* [[I6]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK13-SAME: (i32 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__l122.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__l122([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__l127.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__l127([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__l132.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__l132(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__l137.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__l137([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__l142.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__l142(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__l122
// 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, !llvm.access.group !46
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !46
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP47:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..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, !llvm.access.group !49
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP50:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK13-SAME: ([10 x i32]* 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, !llvm.access.group !52
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !52
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..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, !llvm.access.group !55
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK13-SAME: (i64 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, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !58
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..27 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !58
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..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, !llvm.access.group !61
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK13-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK13-SAME: ([10 x i32]* 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, !llvm.access.group !64
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..31 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !64
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP65:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..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 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK13-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK13-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP68:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK13-SAME: (i64 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, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK13-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !70
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..35 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !70
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP71:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..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 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK13:       omp.dispatch.cond:
// CHECK13-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK13-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK13:       omp.dispatch.body:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP74:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK13:       omp.dispatch.inc:
// CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK13:       omp.dispatch.end:
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK13-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK13-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK13-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@main
// CHECK14-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK14-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK14-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
// CHECK14-NEXT:    [[_TMP9:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_CASTED20:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS23:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES25:%.*]] = alloca [4 x i64], align 8
// CHECK14-NEXT:    [[_TMP26:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_28:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[N_CASTED36:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS38:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS39:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS40:%.*]] = alloca [3 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES41:%.*]] = alloca [3 x i64], align 8
// CHECK14-NEXT:    [[_TMP42:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_43:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_44:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[M_CASTED52:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_CASTED54:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS56:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS57:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS58:%.*]] = alloca [4 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_SIZES59:%.*]] = alloca [4 x i64], align 8
// CHECK14-NEXT:    [[_TMP60:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_61:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_62:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK14-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK14-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK14-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK14-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK14-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK14-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK14-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK14-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK14-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)
// CHECK14-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK14-NEXT:    store i64 [[TMP4]], i64* [[TMP8]], align 8
// CHECK14-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK14-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
// CHECK14-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK14-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP13]], align 8
// CHECK14-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP15]], align 8
// CHECK14-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK14-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP18]], align 8
// CHECK14-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP20]], align 8
// CHECK14-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK14-NEXT:    store i64 [[TMP5]], i64* [[TMP21]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK14-NEXT:    [[TMP29:%.*]] = zext i32 [[ADD]] to i64
// CHECK14-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK14-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP30]], align 4
// CHECK14-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK14-NEXT:    store i32 3, i32* [[TMP31]], align 4
// CHECK14-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP23]], i8*** [[TMP32]], align 8
// CHECK14-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP24]], i8*** [[TMP33]], align 8
// CHECK14-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK14-NEXT:    store i64* [[TMP25]], i64** [[TMP34]], align 8
// CHECK14-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
// CHECK14-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP36]], align 8
// CHECK14-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP37]], align 8
// CHECK14-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK14-NEXT:    store i64 [[TMP29]], i64* [[TMP38]], align 8
// CHECK14-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_l154.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK14-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK14-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK14:       omp_offload.failed:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK14:       omp_offload.cont:
// CHECK14-NEXT:    [[TMP41:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
// CHECK14-NEXT:    store i32 [[TMP41]], i32* [[CONV4]], align 4
// CHECK14-NEXT:    [[TMP42:%.*]] = load i64, i64* [[N_CASTED3]], align 8
// CHECK14-NEXT:    [[TMP43:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP44:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES8]] to i8*
// CHECK14-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)
// CHECK14-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK14-NEXT:    store i64 [[TMP42]], i64* [[TMP46]], align 8
// CHECK14-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK14-NEXT:    store i64 [[TMP42]], i64* [[TMP48]], align 8
// CHECK14-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK14-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP51]], align 8
// CHECK14-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP53]], align 8
// CHECK14-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK14-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP56]], align 8
// CHECK14-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP58]], align 8
// CHECK14-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
// CHECK14-NEXT:    store i64 [[TMP43]], i64* [[TMP59]], align 8
// CHECK14-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK14-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP64:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP64]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK14-NEXT:    [[TMP65:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK14-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP65]], 0
// CHECK14-NEXT:    [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
// CHECK14-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
// CHECK14-NEXT:    store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK14-NEXT:    [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK14-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP66]], 1
// CHECK14-NEXT:    [[TMP67:%.*]] = zext i32 [[ADD15]] to i64
// CHECK14-NEXT:    [[KERNEL_ARGS16:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK14-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK14-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 1
// CHECK14-NEXT:    store i32 3, i32* [[TMP69]], align 4
// CHECK14-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP61]], i8*** [[TMP70]], align 8
// CHECK14-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP62]], i8*** [[TMP71]], align 8
// CHECK14-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 4
// CHECK14-NEXT:    store i64* [[TMP63]], i64** [[TMP72]], align 8
// CHECK14-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP73]], align 8
// CHECK14-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP74]], align 8
// CHECK14-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP75]], align 8
// CHECK14-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 8
// CHECK14-NEXT:    store i64 [[TMP67]], i64* [[TMP76]], align 8
// CHECK14-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_l159.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]])
// CHECK14-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK14-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK14:       omp_offload.failed17:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i64 [[TMP42]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK14:       omp_offload.cont18:
// CHECK14-NEXT:    [[TMP79:%.*]] = load i32, i32* [[M]], align 4
// CHECK14-NEXT:    [[CONV19:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP79]], i32* [[CONV19]], align 4
// CHECK14-NEXT:    [[TMP80:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK14-NEXT:    [[TMP81:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV21:%.*]] = bitcast i64* [[N_CASTED20]] to i32*
// CHECK14-NEXT:    store i32 [[TMP81]], i32* [[CONV21]], align 4
// CHECK14-NEXT:    [[TMP82:%.*]] = load i64, i64* [[N_CASTED20]], align 8
// CHECK14-NEXT:    [[TMP83:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP84:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES25]] to i8*
// CHECK14-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)
// CHECK14-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
// CHECK14-NEXT:    store i64 [[TMP80]], i64* [[TMP86]], align 8
// CHECK14-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
// CHECK14-NEXT:    store i64 [[TMP80]], i64* [[TMP88]], align 8
// CHECK14-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP89]], align 8
// CHECK14-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i64*
// CHECK14-NEXT:    store i64 [[TMP82]], i64* [[TMP91]], align 8
// CHECK14-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i64*
// CHECK14-NEXT:    store i64 [[TMP82]], i64* [[TMP93]], align 8
// CHECK14-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP94]], align 8
// CHECK14-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP96]], align 8
// CHECK14-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP98]], align 8
// CHECK14-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP99]], align 8
// CHECK14-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 3
// CHECK14-NEXT:    [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP101]], align 8
// CHECK14-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 3
// CHECK14-NEXT:    [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP103]], align 8
// CHECK14-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 3
// CHECK14-NEXT:    store i64 [[TMP83]], i64* [[TMP104]], align 8
// CHECK14-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 3
// CHECK14-NEXT:    store i8* null, i8** [[TMP105]], align 8
// CHECK14-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP109:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP109]], i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK14-NEXT:    [[TMP110:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
// CHECK14-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[TMP110]], 0
// CHECK14-NEXT:    [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
// CHECK14-NEXT:    [[SUB31:%.*]] = sub nsw i32 [[DIV30]], 1
// CHECK14-NEXT:    store i32 [[SUB31]], i32* [[DOTCAPTURE_EXPR_28]], align 4
// CHECK14-NEXT:    [[TMP111:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_28]], align 4
// CHECK14-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP111]], 1
// CHECK14-NEXT:    [[TMP112:%.*]] = zext i32 [[ADD32]] to i64
// CHECK14-NEXT:    [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK14-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP113]], align 4
// CHECK14-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 1
// CHECK14-NEXT:    store i32 4, i32* [[TMP114]], align 4
// CHECK14-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP106]], i8*** [[TMP115]], align 8
// CHECK14-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP107]], i8*** [[TMP116]], align 8
// CHECK14-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 4
// CHECK14-NEXT:    store i64* [[TMP108]], i64** [[TMP117]], align 8
// CHECK14-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP118]], align 8
// CHECK14-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP119]], align 8
// CHECK14-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP120]], align 8
// CHECK14-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 8
// CHECK14-NEXT:    store i64 [[TMP112]], i64* [[TMP121]], align 8
// CHECK14-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_l164.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]])
// CHECK14-NEXT:    [[TMP123:%.*]] = icmp ne i32 [[TMP122]], 0
// CHECK14-NEXT:    br i1 [[TMP123]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
// CHECK14:       omp_offload.failed34:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i64 [[TMP80]], i64 [[TMP82]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT35]]
// CHECK14:       omp_offload.cont35:
// CHECK14-NEXT:    [[TMP124:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV37:%.*]] = bitcast i64* [[N_CASTED36]] to i32*
// CHECK14-NEXT:    store i32 [[TMP124]], i32* [[CONV37]], align 4
// CHECK14-NEXT:    [[TMP125:%.*]] = load i64, i64* [[N_CASTED36]], align 8
// CHECK14-NEXT:    [[TMP126:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP127:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES41]] to i8*
// CHECK14-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)
// CHECK14-NEXT:    [[TMP128:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP129:%.*]] = bitcast i8** [[TMP128]] to i64*
// CHECK14-NEXT:    store i64 [[TMP125]], i64* [[TMP129]], align 8
// CHECK14-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i64*
// CHECK14-NEXT:    store i64 [[TMP125]], i64* [[TMP131]], align 8
// CHECK14-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP132]], align 8
// CHECK14-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP134:%.*]] = bitcast i8** [[TMP133]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP134]], align 8
// CHECK14-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP136]], align 8
// CHECK14-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP137]], align 8
// CHECK14-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP139]], align 8
// CHECK14-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP141]], align 8
// CHECK14-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES41]], i32 0, i32 2
// CHECK14-NEXT:    store i64 [[TMP126]], i64* [[TMP142]], align 8
// CHECK14-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS40]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP143]], align 8
// CHECK14-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS38]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS39]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES41]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP147:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP147]], i32* [[DOTCAPTURE_EXPR_43]], align 4
// CHECK14-NEXT:    [[TMP148:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_43]], align 4
// CHECK14-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[TMP148]], 0
// CHECK14-NEXT:    [[DIV46:%.*]] = sdiv i32 [[SUB45]], 1
// CHECK14-NEXT:    [[SUB47:%.*]] = sub nsw i32 [[DIV46]], 1
// CHECK14-NEXT:    store i32 [[SUB47]], i32* [[DOTCAPTURE_EXPR_44]], align 4
// CHECK14-NEXT:    [[TMP149:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_44]], align 4
// CHECK14-NEXT:    [[ADD48:%.*]] = add nsw i32 [[TMP149]], 1
// CHECK14-NEXT:    [[TMP150:%.*]] = zext i32 [[ADD48]] to i64
// CHECK14-NEXT:    [[KERNEL_ARGS49:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK14-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP151]], align 4
// CHECK14-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 1
// CHECK14-NEXT:    store i32 3, i32* [[TMP152]], align 4
// CHECK14-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP144]], i8*** [[TMP153]], align 8
// CHECK14-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP145]], i8*** [[TMP154]], align 8
// CHECK14-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 4
// CHECK14-NEXT:    store i64* [[TMP146]], i64** [[TMP155]], align 8
// CHECK14-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP156]], align 8
// CHECK14-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP157]], align 8
// CHECK14-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP158]], align 8
// CHECK14-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]], i32 0, i32 8
// CHECK14-NEXT:    store i64 [[TMP150]], i64* [[TMP159]], align 8
// CHECK14-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_l169.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS49]])
// CHECK14-NEXT:    [[TMP161:%.*]] = icmp ne i32 [[TMP160]], 0
// CHECK14-NEXT:    br i1 [[TMP161]], label [[OMP_OFFLOAD_FAILED50:%.*]], label [[OMP_OFFLOAD_CONT51:%.*]]
// CHECK14:       omp_offload.failed50:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i64 [[TMP125]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT51]]
// CHECK14:       omp_offload.cont51:
// CHECK14-NEXT:    [[TMP162:%.*]] = load i32, i32* [[M]], align 4
// CHECK14-NEXT:    [[CONV53:%.*]] = bitcast i64* [[M_CASTED52]] to i32*
// CHECK14-NEXT:    store i32 [[TMP162]], i32* [[CONV53]], align 4
// CHECK14-NEXT:    [[TMP163:%.*]] = load i64, i64* [[M_CASTED52]], align 8
// CHECK14-NEXT:    [[TMP164:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CONV55:%.*]] = bitcast i64* [[N_CASTED54]] to i32*
// CHECK14-NEXT:    store i32 [[TMP164]], i32* [[CONV55]], align 4
// CHECK14-NEXT:    [[TMP165:%.*]] = load i64, i64* [[N_CASTED54]], align 8
// CHECK14-NEXT:    [[TMP166:%.*]] = mul nuw i64 [[TMP1]], 4
// CHECK14-NEXT:    [[TMP167:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES59]] to i8*
// CHECK14-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)
// CHECK14-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP169:%.*]] = bitcast i8** [[TMP168]] to i64*
// CHECK14-NEXT:    store i64 [[TMP163]], i64* [[TMP169]], align 8
// CHECK14-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP171:%.*]] = bitcast i8** [[TMP170]] to i64*
// CHECK14-NEXT:    store i64 [[TMP163]], i64* [[TMP171]], align 8
// CHECK14-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP172]], align 8
// CHECK14-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP174:%.*]] = bitcast i8** [[TMP173]] to i64*
// CHECK14-NEXT:    store i64 [[TMP165]], i64* [[TMP174]], align 8
// CHECK14-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP176:%.*]] = bitcast i8** [[TMP175]] to i64*
// CHECK14-NEXT:    store i64 [[TMP165]], i64* [[TMP176]], align 8
// CHECK14-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP177]], align 8
// CHECK14-NEXT:    [[TMP178:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP179:%.*]] = bitcast i8** [[TMP178]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP179]], align 8
// CHECK14-NEXT:    [[TMP180:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 2
// CHECK14-NEXT:    [[TMP181:%.*]] = bitcast i8** [[TMP180]] to i64*
// CHECK14-NEXT:    store i64 [[TMP1]], i64* [[TMP181]], align 8
// CHECK14-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 2
// CHECK14-NEXT:    store i8* null, i8** [[TMP182]], align 8
// CHECK14-NEXT:    [[TMP183:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 3
// CHECK14-NEXT:    [[TMP184:%.*]] = bitcast i8** [[TMP183]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP184]], align 8
// CHECK14-NEXT:    [[TMP185:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 3
// CHECK14-NEXT:    [[TMP186:%.*]] = bitcast i8** [[TMP185]] to i32**
// CHECK14-NEXT:    store i32* [[VLA]], i32** [[TMP186]], align 8
// CHECK14-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES59]], i32 0, i32 3
// CHECK14-NEXT:    store i64 [[TMP166]], i64* [[TMP187]], align 8
// CHECK14-NEXT:    [[TMP188:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS58]], i64 0, i64 3
// CHECK14-NEXT:    store i8* null, i8** [[TMP188]], align 8
// CHECK14-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS56]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP190:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS57]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES59]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP192:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    store i32 [[TMP192]], i32* [[DOTCAPTURE_EXPR_61]], align 4
// CHECK14-NEXT:    [[TMP193:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_61]], align 4
// CHECK14-NEXT:    [[SUB63:%.*]] = sub nsw i32 [[TMP193]], 0
// CHECK14-NEXT:    [[DIV64:%.*]] = sdiv i32 [[SUB63]], 1
// CHECK14-NEXT:    [[SUB65:%.*]] = sub nsw i32 [[DIV64]], 1
// CHECK14-NEXT:    store i32 [[SUB65]], i32* [[DOTCAPTURE_EXPR_62]], align 4
// CHECK14-NEXT:    [[TMP194:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_62]], align 4
// CHECK14-NEXT:    [[ADD66:%.*]] = add nsw i32 [[TMP194]], 1
// CHECK14-NEXT:    [[TMP195:%.*]] = zext i32 [[ADD66]] to i64
// CHECK14-NEXT:    [[KERNEL_ARGS67:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK14-NEXT:    [[TMP196:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP196]], align 4
// CHECK14-NEXT:    [[TMP197:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 1
// CHECK14-NEXT:    store i32 4, i32* [[TMP197]], align 4
// CHECK14-NEXT:    [[TMP198:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP189]], i8*** [[TMP198]], align 8
// CHECK14-NEXT:    [[TMP199:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP190]], i8*** [[TMP199]], align 8
// CHECK14-NEXT:    [[TMP200:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 4
// CHECK14-NEXT:    store i64* [[TMP191]], i64** [[TMP200]], align 8
// CHECK14-NEXT:    [[TMP201:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP201]], align 8
// CHECK14-NEXT:    [[TMP202:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP202]], align 8
// CHECK14-NEXT:    [[TMP203:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP203]], align 8
// CHECK14-NEXT:    [[TMP204:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]], i32 0, i32 8
// CHECK14-NEXT:    store i64 [[TMP195]], i64* [[TMP204]], align 8
// CHECK14-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_l174.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS67]])
// CHECK14-NEXT:    [[TMP206:%.*]] = icmp ne i32 [[TMP205]], 0
// CHECK14-NEXT:    br i1 [[TMP206]], label [[OMP_OFFLOAD_FAILED68:%.*]], label [[OMP_OFFLOAD_CONT69:%.*]]
// CHECK14:       omp_offload.failed68:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i64 [[TMP163]], i64 [[TMP165]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT69]]
// CHECK14:       omp_offload.cont69:
// CHECK14-NEXT:    [[TMP207:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK14-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP207]])
// CHECK14-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK14-NEXT:    [[TMP208:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK14-NEXT:    call void @llvm.stackrestore(i8* [[TMP208]])
// CHECK14-NEXT:    [[TMP209:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK14-NEXT:    ret i32 [[TMP209]]
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK14-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK14-SAME: (i32* noalias 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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !13
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK14-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK14-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK14-SAME: (i32* noalias 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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !17
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK14-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK14-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK14-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK14-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK14-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK14-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK14-SAME: (i32* noalias 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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !22
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK14-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK14-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK14-SAME: (i32* noalias 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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK14-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK14-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK14-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK14-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK14-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK14-SAME: (i64 noundef [[M:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK14-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP21]], i32* [[CONV7]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !28
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]]), !llvm.access.group !28
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
// CHECK14-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK14:       cond.true12:
// CHECK14-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    br label [[COND_END14:%.*]]
// CHECK14:       cond.false13:
// CHECK14-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    br label [[COND_END14]]
// CHECK14:       cond.end14:
// CHECK14-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
// CHECK14-NEXT:    store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
// CHECK14-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK14-NEXT:    br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK14-NEXT:    [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
// CHECK14-NEXT:    [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD18]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK14-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK14-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK14-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK14-NEXT:    [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
// CHECK14-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
// CHECK14-NEXT:    [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
// CHECK14-NEXT:    store i32 [[ADD13]], i32* [[I6]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK14-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !34
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK14-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK14-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK14-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK14-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK14-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK14-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK14-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK14-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK14-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK14-SAME: (i64 noundef [[M:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK14-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP20]], i32* [[CONV7]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !40
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]]), !llvm.access.group !40
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK14-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
// CHECK14-NEXT:    br i1 [[TMP27]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP28]], 0
// CHECK14-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK14-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK14-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK14-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK14:       omp.precond.then:
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK14-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK14-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK14-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK14-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK14-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK14-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK14-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK14-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
// CHECK14-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
// CHECK14-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK14-NEXT:    store i32 [[ADD12]], i32* [[I6]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK14:       omp.precond.end:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK14-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK14-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [1 x i8*], align 8
// CHECK14-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[M_CASTED22:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [2 x i8*], align 8
// CHECK14-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK14-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK14-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
// CHECK14-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK14-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK14-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK14-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK14-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK14-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK14-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 8
// CHECK14-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 8
// CHECK14-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK14-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP11]], align 8
// CHECK14-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP12]], align 8
// CHECK14-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP13]], align 8
// CHECK14-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP14]], align 8
// CHECK14-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK14-NEXT:    store i64 10, i64* [[TMP15]], align 8
// CHECK14-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__l122.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK14-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK14-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK14:       omp_offload.failed:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK14:       omp_offload.cont:
// CHECK14-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 8
// CHECK14-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 8
// CHECK14-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK14-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK14-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK14-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK14-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
// CHECK14-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK14-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 8
// CHECK14-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 8
// CHECK14-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
// CHECK14-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.24, i32 0, i32 0), i64** [[TMP29]], align 8
// CHECK14-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.25, i32 0, i32 0), i64** [[TMP30]], align 8
// CHECK14-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP31]], align 8
// CHECK14-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP32]], align 8
// CHECK14-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
// CHECK14-NEXT:    store i64 10, i64* [[TMP33]], align 8
// CHECK14-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__l127.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
// CHECK14-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK14-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK14:       omp_offload.failed6:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK14:       omp_offload.cont7:
// CHECK14-NEXT:    [[TMP36:%.*]] = load i32, i32* [[M]], align 4
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP36]], i32* [[CONV]], align 4
// CHECK14-NEXT:    [[TMP37:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK14-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK14-NEXT:    store i64 [[TMP37]], i64* [[TMP39]], align 8
// CHECK14-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
// CHECK14-NEXT:    store i64 [[TMP37]], i64* [[TMP41]], align 8
// CHECK14-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP42]], align 8
// CHECK14-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 8
// CHECK14-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 8
// CHECK14-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP47]], align 8
// CHECK14-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK14-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK14-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP50]], align 4
// CHECK14-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
// CHECK14-NEXT:    store i32 2, i32* [[TMP51]], align 4
// CHECK14-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP48]], i8*** [[TMP52]], align 8
// CHECK14-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP49]], i8*** [[TMP53]], align 8
// CHECK14-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
// CHECK14-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.28, i32 0, i32 0), i64** [[TMP54]], align 8
// CHECK14-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.29, i32 0, i32 0), i64** [[TMP55]], align 8
// CHECK14-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP56]], align 8
// CHECK14-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP57]], align 8
// CHECK14-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
// CHECK14-NEXT:    store i64 10, i64* [[TMP58]], align 8
// CHECK14-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__l132.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
// CHECK14-NEXT:    [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
// CHECK14-NEXT:    br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
// CHECK14:       omp_offload.failed13:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132(i64 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
// CHECK14:       omp_offload.cont14:
// CHECK14-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP62]], align 8
// CHECK14-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP64]], align 8
// CHECK14-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS17]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP65]], align 8
// CHECK14-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
// CHECK14-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK14-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK14-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 1
// CHECK14-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK14-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP66]], i8*** [[TMP70]], align 8
// CHECK14-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP67]], i8*** [[TMP71]], align 8
// CHECK14-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 4
// CHECK14-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.32, i32 0, i32 0), i64** [[TMP72]], align 8
// CHECK14-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.33, i32 0, i32 0), i64** [[TMP73]], align 8
// CHECK14-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP74]], align 8
// CHECK14-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP75]], align 8
// CHECK14-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 8
// CHECK14-NEXT:    store i64 10, i64* [[TMP76]], align 8
// CHECK14-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__l137.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]])
// CHECK14-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK14-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
// CHECK14:       omp_offload.failed20:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137([10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
// CHECK14:       omp_offload.cont21:
// CHECK14-NEXT:    [[TMP79:%.*]] = load i32, i32* [[M]], align 4
// CHECK14-NEXT:    [[CONV23:%.*]] = bitcast i64* [[M_CASTED22]] to i32*
// CHECK14-NEXT:    store i32 [[TMP79]], i32* [[CONV23]], align 4
// CHECK14-NEXT:    [[TMP80:%.*]] = load i64, i64* [[M_CASTED22]], align 8
// CHECK14-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
// CHECK14-NEXT:    store i64 [[TMP80]], i64* [[TMP82]], align 8
// CHECK14-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP84:%.*]] = bitcast i8** [[TMP83]] to i64*
// CHECK14-NEXT:    store i64 [[TMP80]], i64* [[TMP84]], align 8
// CHECK14-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
// CHECK14-NEXT:    store i8* null, i8** [[TMP85]], align 8
// CHECK14-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP87]], align 8
// CHECK14-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 1
// CHECK14-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to [10 x i32]**
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP89]], align 8
// CHECK14-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 1
// CHECK14-NEXT:    store i8* null, i8** [[TMP90]], align 8
// CHECK14-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK14-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK14-NEXT:    [[KERNEL_ARGS28:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK14-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 0
// CHECK14-NEXT:    store i32 1, i32* [[TMP93]], align 4
// CHECK14-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 1
// CHECK14-NEXT:    store i32 2, i32* [[TMP94]], align 4
// CHECK14-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 2
// CHECK14-NEXT:    store i8** [[TMP91]], i8*** [[TMP95]], align 8
// CHECK14-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 3
// CHECK14-NEXT:    store i8** [[TMP92]], i8*** [[TMP96]], align 8
// CHECK14-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 4
// CHECK14-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.36, i32 0, i32 0), i64** [[TMP97]], align 8
// CHECK14-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 5
// CHECK14-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.37, i32 0, i32 0), i64** [[TMP98]], align 8
// CHECK14-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 6
// CHECK14-NEXT:    store i8** null, i8*** [[TMP99]], align 8
// CHECK14-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 7
// CHECK14-NEXT:    store i8** null, i8*** [[TMP100]], align 8
// CHECK14-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 8
// CHECK14-NEXT:    store i64 10, i64* [[TMP101]], align 8
// CHECK14-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__l142.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]])
// CHECK14-NEXT:    [[TMP103:%.*]] = icmp ne i32 [[TMP102]], 0
// CHECK14-NEXT:    br i1 [[TMP103]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
// CHECK14:       omp_offload.failed29:
// CHECK14-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i64 [[TMP80]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK14-NEXT:    br label [[OMP_OFFLOAD_CONT30]]
// CHECK14:       omp_offload.cont30:
// CHECK14-NEXT:    ret i32 0
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// CHECK14-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK14-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !46
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP47:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK14-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..19
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK14-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK14-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !49
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP50:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK14-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK14-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..22 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..22
// CHECK14-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !52
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !52
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK14-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..23
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK14-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK14-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK14-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK14-SAME: (i64 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK14-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..26 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..26
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !58
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..27 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !58
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK14-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..27
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK14-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK14-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..30 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..30
// CHECK14-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..31 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !64
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !64
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP65:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK14-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..31
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK14-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK14-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK14-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK14-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !67
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP68:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK14-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK14-SAME: (i64 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK14-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..34 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..34
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK14:       cond.true:
// CHECK14-NEXT:    br label [[COND_END:%.*]]
// CHECK14:       cond.false:
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    br label [[COND_END]]
// CHECK14:       cond.end:
// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK14-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK14-NEXT:    store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !70
// CHECK14-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..35 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !70
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !70
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP71:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK14:       omp.loop.exit:
// CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK14-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..35
// CHECK14-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]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
// CHECK14-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK14-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK14-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK14-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK14-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK14:       omp.dispatch.cond:
// CHECK14-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK14-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK14-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK14:       omp.dispatch.body:
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK14-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK14:       omp.inner.for.cond:
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK14:       omp.inner.for.body:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK14:       omp.body.continue:
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK14:       omp.inner.for.inc:
// CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK14-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !73
// CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP74:![0-9]+]]
// CHECK14:       omp.inner.for.end:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK14:       omp.dispatch.inc:
// CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK14:       omp.dispatch.end:
// CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK14-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK14-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK14:       .omp.final.then:
// CHECK14-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK14:       .omp.final.done:
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK14-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK14-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@main
// CHECK17-SAME: (i32 noundef [[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 4
// CHECK17-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
// CHECK17-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED18:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES22:%.*]] = alloca [4 x i64], align 4
// CHECK17-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED33:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS34:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS35:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS36:%.*]] = alloca [3 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES37:%.*]] = alloca [3 x i64], align 4
// CHECK17-NEXT:    [[_TMP38:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED48:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_CASTED49:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS50:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS51:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS52:%.*]] = alloca [4 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES53:%.*]] = alloca [4 x i64], align 4
// CHECK17-NEXT:    [[_TMP54:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_55:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_56:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK17-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK17-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK17-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK17-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK17-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK17-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[N_CASTED]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
// CHECK17-NEXT:    [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK17-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)
// CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[TMP8]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[TMP10]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP13]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP15]], align 4
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP16]], align 4
// 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 4
// 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 4
// 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 4
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP22]], align 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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_l154.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_l154(i32 [[TMP3]], i32 [[TMP0]], 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:    store i32 [[TMP41]], i32* [[N_CASTED3]], align 4
// CHECK17-NEXT:    [[TMP42:%.*]] = load i32, i32* [[N_CASTED3]], align 4
// CHECK17-NEXT:    [[TMP43:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP44:%.*]] = sext i32 [[TMP43]] to i64
// CHECK17-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES7]] to i8*
// CHECK17-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)
// CHECK17-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
// CHECK17-NEXT:    store i32 [[TMP42]], i32* [[TMP47]], align 4
// CHECK17-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
// CHECK17-NEXT:    store i32 [[TMP42]], i32* [[TMP49]], align 4
// CHECK17-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP50]], align 4
// CHECK17-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP52]], align 4
// CHECK17-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP54:%.*]] = bitcast i8** [[TMP53]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP54]], align 4
// CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP55]], align 4
// CHECK17-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP57]], align 4
// CHECK17-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP59]], align 4
// CHECK17-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP44]], i64* [[TMP60]], align 4
// CHECK17-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP61]], align 4
// CHECK17-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP65:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP65]], i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK17-NEXT:    [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
// CHECK17-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP66]], 0
// CHECK17-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
// CHECK17-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
// CHECK17-NEXT:    store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK17-NEXT:    [[TMP67:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
// CHECK17-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP67]], 1
// CHECK17-NEXT:    [[TMP68:%.*]] = zext i32 [[ADD14]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK17-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
// CHECK17-NEXT:    store i32 3, i32* [[TMP70]], align 4
// CHECK17-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP62]], i8*** [[TMP71]], align 4
// CHECK17-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP63]], i8*** [[TMP72]], align 4
// CHECK17-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP64]], i64** [[TMP73]], align 4
// CHECK17-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP74]], align 4
// CHECK17-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP75]], align 4
// CHECK17-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP76]], align 4
// CHECK17-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP68]], i64* [[TMP77]], align 8
// CHECK17-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_l159.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
// CHECK17-NEXT:    [[TMP79:%.*]] = icmp ne i32 [[TMP78]], 0
// CHECK17-NEXT:    br i1 [[TMP79]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK17:       omp_offload.failed16:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159(i32 [[TMP42]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK17:       omp_offload.cont17:
// CHECK17-NEXT:    [[TMP80:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    store i32 [[TMP80]], i32* [[M_CASTED]], align 4
// CHECK17-NEXT:    [[TMP81:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK17-NEXT:    [[TMP82:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP82]], i32* [[N_CASTED18]], align 4
// CHECK17-NEXT:    [[TMP83:%.*]] = load i32, i32* [[N_CASTED18]], align 4
// CHECK17-NEXT:    [[TMP84:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP85:%.*]] = sext i32 [[TMP84]] to i64
// CHECK17-NEXT:    [[TMP86:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES22]] to i8*
// CHECK17-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)
// CHECK17-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
// CHECK17-NEXT:    store i32 [[TMP81]], i32* [[TMP88]], align 4
// CHECK17-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP90:%.*]] = bitcast i8** [[TMP89]] to i32*
// CHECK17-NEXT:    store i32 [[TMP81]], i32* [[TMP90]], align 4
// CHECK17-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP91]], align 4
// CHECK17-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i32*
// CHECK17-NEXT:    store i32 [[TMP83]], i32* [[TMP93]], align 4
// CHECK17-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP95:%.*]] = bitcast i8** [[TMP94]] to i32*
// CHECK17-NEXT:    store i32 [[TMP83]], i32* [[TMP95]], align 4
// CHECK17-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP96]], align 4
// CHECK17-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP98]], align 4
// CHECK17-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP100]], align 4
// CHECK17-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP101]], align 4
// CHECK17-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP103]], align 4
// CHECK17-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP105]], align 4
// CHECK17-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP85]], i64* [[TMP106]], align 4
// CHECK17-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP107]], align 4
// CHECK17-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP111:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP111]], i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK17-NEXT:    [[TMP112:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
// CHECK17-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[TMP112]], 0
// CHECK17-NEXT:    [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
// CHECK17-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[DIV27]], 1
// CHECK17-NEXT:    store i32 [[SUB28]], i32* [[DOTCAPTURE_EXPR_25]], align 4
// CHECK17-NEXT:    [[TMP113:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
// CHECK17-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP113]], 1
// CHECK17-NEXT:    [[TMP114:%.*]] = zext i32 [[ADD29]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS30:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP115]], align 4
// CHECK17-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 1
// CHECK17-NEXT:    store i32 4, i32* [[TMP116]], align 4
// CHECK17-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP108]], i8*** [[TMP117]], align 4
// CHECK17-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP109]], i8*** [[TMP118]], align 4
// CHECK17-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP110]], i64** [[TMP119]], align 4
// CHECK17-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP120]], align 4
// CHECK17-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP121]], align 4
// CHECK17-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP122]], align 4
// CHECK17-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP114]], i64* [[TMP123]], align 8
// CHECK17-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_l164.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]])
// CHECK17-NEXT:    [[TMP125:%.*]] = icmp ne i32 [[TMP124]], 0
// CHECK17-NEXT:    br i1 [[TMP125]], label [[OMP_OFFLOAD_FAILED31:%.*]], label [[OMP_OFFLOAD_CONT32:%.*]]
// CHECK17:       omp_offload.failed31:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164(i32 [[TMP81]], i32 [[TMP83]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT32]]
// CHECK17:       omp_offload.cont32:
// CHECK17-NEXT:    [[TMP126:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP126]], i32* [[N_CASTED33]], align 4
// CHECK17-NEXT:    [[TMP127:%.*]] = load i32, i32* [[N_CASTED33]], align 4
// CHECK17-NEXT:    [[TMP128:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP129:%.*]] = sext i32 [[TMP128]] to i64
// CHECK17-NEXT:    [[TMP130:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES37]] to i8*
// CHECK17-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)
// CHECK17-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP132:%.*]] = bitcast i8** [[TMP131]] to i32*
// CHECK17-NEXT:    store i32 [[TMP127]], i32* [[TMP132]], align 4
// CHECK17-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP134:%.*]] = bitcast i8** [[TMP133]] to i32*
// CHECK17-NEXT:    store i32 [[TMP127]], i32* [[TMP134]], align 4
// CHECK17-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP135]], align 4
// CHECK17-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP137]], align 4
// CHECK17-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP139]], align 4
// CHECK17-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP140]], align 4
// CHECK17-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP142]], align 4
// CHECK17-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP144]], align 4
// CHECK17-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES37]], i32 0, i32 2
// CHECK17-NEXT:    store i64 [[TMP129]], i64* [[TMP145]], align 4
// CHECK17-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS36]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP146]], align 4
// CHECK17-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS34]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS35]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP149:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES37]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP150:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP150]], i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK17-NEXT:    [[TMP151:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK17-NEXT:    [[SUB41:%.*]] = sub nsw i32 [[TMP151]], 0
// CHECK17-NEXT:    [[DIV42:%.*]] = sdiv i32 [[SUB41]], 1
// CHECK17-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[DIV42]], 1
// CHECK17-NEXT:    store i32 [[SUB43]], i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK17-NEXT:    [[TMP152:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK17-NEXT:    [[ADD44:%.*]] = add nsw i32 [[TMP152]], 1
// CHECK17-NEXT:    [[TMP153:%.*]] = zext i32 [[ADD44]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS45:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP154]], align 4
// CHECK17-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 1
// CHECK17-NEXT:    store i32 3, i32* [[TMP155]], align 4
// CHECK17-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP147]], i8*** [[TMP156]], align 4
// CHECK17-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP148]], i8*** [[TMP157]], align 4
// CHECK17-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP149]], i64** [[TMP158]], align 4
// CHECK17-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP159]], align 4
// CHECK17-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP160]], align 4
// CHECK17-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP161]], align 4
// CHECK17-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP153]], i64* [[TMP162]], align 8
// CHECK17-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_l169.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS45]])
// CHECK17-NEXT:    [[TMP164:%.*]] = icmp ne i32 [[TMP163]], 0
// CHECK17-NEXT:    br i1 [[TMP164]], label [[OMP_OFFLOAD_FAILED46:%.*]], label [[OMP_OFFLOAD_CONT47:%.*]]
// CHECK17:       omp_offload.failed46:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169(i32 [[TMP127]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT47]]
// CHECK17:       omp_offload.cont47:
// CHECK17-NEXT:    [[TMP165:%.*]] = load i32, i32* [[M]], align 4
// CHECK17-NEXT:    store i32 [[TMP165]], i32* [[M_CASTED48]], align 4
// CHECK17-NEXT:    [[TMP166:%.*]] = load i32, i32* [[M_CASTED48]], align 4
// CHECK17-NEXT:    [[TMP167:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP167]], i32* [[N_CASTED49]], align 4
// CHECK17-NEXT:    [[TMP168:%.*]] = load i32, i32* [[N_CASTED49]], align 4
// CHECK17-NEXT:    [[TMP169:%.*]] = mul nuw i32 [[TMP0]], 4
// CHECK17-NEXT:    [[TMP170:%.*]] = sext i32 [[TMP169]] to i64
// CHECK17-NEXT:    [[TMP171:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES53]] to i8*
// CHECK17-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)
// CHECK17-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP173:%.*]] = bitcast i8** [[TMP172]] to i32*
// CHECK17-NEXT:    store i32 [[TMP166]], i32* [[TMP173]], align 4
// CHECK17-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP175:%.*]] = bitcast i8** [[TMP174]] to i32*
// CHECK17-NEXT:    store i32 [[TMP166]], i32* [[TMP175]], align 4
// CHECK17-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP176]], align 4
// CHECK17-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP178:%.*]] = bitcast i8** [[TMP177]] to i32*
// CHECK17-NEXT:    store i32 [[TMP168]], i32* [[TMP178]], align 4
// CHECK17-NEXT:    [[TMP179:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP180:%.*]] = bitcast i8** [[TMP179]] to i32*
// CHECK17-NEXT:    store i32 [[TMP168]], i32* [[TMP180]], align 4
// CHECK17-NEXT:    [[TMP181:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP181]], align 4
// CHECK17-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP183:%.*]] = bitcast i8** [[TMP182]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP183]], align 4
// CHECK17-NEXT:    [[TMP184:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP185:%.*]] = bitcast i8** [[TMP184]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[TMP185]], align 4
// CHECK17-NEXT:    [[TMP186:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP186]], align 4
// CHECK17-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP188:%.*]] = bitcast i8** [[TMP187]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP188]], align 4
// CHECK17-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP190:%.*]] = bitcast i8** [[TMP189]] to i32**
// CHECK17-NEXT:    store i32* [[VLA]], i32** [[TMP190]], align 4
// CHECK17-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES53]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP170]], i64* [[TMP191]], align 4
// CHECK17-NEXT:    [[TMP192:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP192]], align 4
// CHECK17-NEXT:    [[TMP193:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP194:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP195:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES53]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP196:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    store i32 [[TMP196]], i32* [[DOTCAPTURE_EXPR_55]], align 4
// CHECK17-NEXT:    [[TMP197:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_55]], align 4
// CHECK17-NEXT:    [[SUB57:%.*]] = sub nsw i32 [[TMP197]], 0
// CHECK17-NEXT:    [[DIV58:%.*]] = sdiv i32 [[SUB57]], 1
// CHECK17-NEXT:    [[SUB59:%.*]] = sub nsw i32 [[DIV58]], 1
// CHECK17-NEXT:    store i32 [[SUB59]], i32* [[DOTCAPTURE_EXPR_56]], align 4
// CHECK17-NEXT:    [[TMP198:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_56]], align 4
// CHECK17-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP198]], 1
// CHECK17-NEXT:    [[TMP199:%.*]] = zext i32 [[ADD60]] to i64
// CHECK17-NEXT:    [[KERNEL_ARGS61:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP200:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 0
// CHECK17-NEXT:    store i32 1, i32* [[TMP200]], align 4
// CHECK17-NEXT:    [[TMP201:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 1
// CHECK17-NEXT:    store i32 4, i32* [[TMP201]], align 4
// CHECK17-NEXT:    [[TMP202:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP193]], i8*** [[TMP202]], align 4
// CHECK17-NEXT:    [[TMP203:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP194]], i8*** [[TMP203]], align 4
// CHECK17-NEXT:    [[TMP204:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 4
// CHECK17-NEXT:    store i64* [[TMP195]], i64** [[TMP204]], align 4
// CHECK17-NEXT:    [[TMP205:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 5
// CHECK17-NEXT:    store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP205]], align 4
// CHECK17-NEXT:    [[TMP206:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP206]], align 4
// CHECK17-NEXT:    [[TMP207:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP207]], align 4
// CHECK17-NEXT:    [[TMP208:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]], i32 0, i32 8
// CHECK17-NEXT:    store i64 [[TMP199]], i64* [[TMP208]], align 8
// CHECK17-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_l174.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS61]])
// CHECK17-NEXT:    [[TMP210:%.*]] = icmp ne i32 [[TMP209]], 0
// CHECK17-NEXT:    br i1 [[TMP210]], label [[OMP_OFFLOAD_FAILED62:%.*]], label [[OMP_OFFLOAD_CONT63:%.*]]
// CHECK17:       omp_offload.failed62:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174(i32 [[TMP166]], i32 [[TMP168]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT63]]
// CHECK17:       omp_offload.cont63:
// CHECK17-NEXT:    [[TMP211:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK17-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP211]])
// CHECK17-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK17-NEXT:    [[TMP212:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP212]])
// CHECK17-NEXT:    [[TMP213:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK17-NEXT:    ret i32 [[TMP213]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l154
// CHECK17-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK17-SAME: (i32* noalias 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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !14
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK17-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK17-SAME: (i32* noalias 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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK17-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK17-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK17-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK17-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK17-SAME: (i32* noalias 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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !23
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK17-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK17-SAME: (i32* noalias 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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK17-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK17-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK17-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK17-SAME: (i32 noundef [[M:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[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:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK17-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]]), !llvm.access.group !29
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
// CHECK17-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK17:       cond.true11:
// CHECK17-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    br label [[COND_END13:%.*]]
// CHECK17:       cond.false12:
// CHECK17-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    br label [[COND_END13]]
// CHECK17:       cond.end13:
// CHECK17-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
// CHECK17-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
// CHECK17-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK17-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK17-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
// CHECK17-NEXT:    [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD17]], i32* [[I4]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK17-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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK17-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK17-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK17-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK17-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK17-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK17-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[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:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !35
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK17-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK17-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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK17-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK17-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK17-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK17-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK17-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK17-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK17-SAME: (i32 noundef [[M:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[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:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]]), !llvm.access.group !41
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK17-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK17-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[I4]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK17-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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK17-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK17-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK17:       omp.precond.then:
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK17-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK17-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK17-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK17-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK17-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK17-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK17-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK17-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK17-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK17:       omp.precond.end:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK17-SAME: (i32 noundef [[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 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [1 x i8*], align 4
// CHECK17-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[M_CASTED22:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS23:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS24:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS25:%.*]] = alloca [2 x i8*], align 4
// CHECK17-NEXT:    [[_TMP26:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK17-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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__l122.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__l122([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 4
// 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 4
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP22]], align 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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__l127.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__l127([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:    store i32 [[TMP36]], i32* [[M_CASTED]], align 4
// CHECK17-NEXT:    [[TMP37:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i32*
// CHECK17-NEXT:    store i32 [[TMP37]], i32* [[TMP39]], align 4
// CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i32*
// CHECK17-NEXT:    store i32 [[TMP37]], i32* [[TMP41]], align 4
// CHECK17-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP42]], align 4
// 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 4
// 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 4
// CHECK17-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP47]], align 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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__l132.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__l132(i32 [[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 4
// 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 4
// CHECK17-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS17]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP65]], align 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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 4
// 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__l137.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__l137([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:    store i32 [[TMP79]], i32* [[M_CASTED22]], align 4
// CHECK17-NEXT:    [[TMP80:%.*]] = load i32, i32* [[M_CASTED22]], align 4
// CHECK17-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
// CHECK17-NEXT:    store i32 [[TMP80]], i32* [[TMP82]], align 4
// CHECK17-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP84:%.*]] = bitcast i8** [[TMP83]] to i32*
// CHECK17-NEXT:    store i32 [[TMP80]], i32* [[TMP84]], align 4
// CHECK17-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP85]], align 4
// CHECK17-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], 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 4
// CHECK17-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], 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 4
// CHECK17-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP90]], align 4
// CHECK17-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
// CHECK17-NEXT:    [[KERNEL_ARGS27:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK17-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], 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_ARGS27]], 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_ARGS27]], i32 0, i32 2
// CHECK17-NEXT:    store i8** [[TMP91]], i8*** [[TMP95]], align 4
// CHECK17-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 3
// CHECK17-NEXT:    store i8** [[TMP92]], i8*** [[TMP96]], align 4
// CHECK17-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], 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 4
// CHECK17-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], 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 4
// CHECK17-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 6
// CHECK17-NEXT:    store i8** null, i8*** [[TMP99]], align 4
// CHECK17-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], i32 0, i32 7
// CHECK17-NEXT:    store i8** null, i8*** [[TMP100]], align 4
// CHECK17-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]], 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__l142.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS27]])
// CHECK17-NEXT:    [[TMP103:%.*]] = icmp ne i32 [[TMP102]], 0
// CHECK17-NEXT:    br i1 [[TMP103]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
// CHECK17:       omp_offload.failed28:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142(i32 [[TMP80]], [10 x i32]* [[A]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
// CHECK17:       omp_offload.cont29:
// CHECK17-NEXT:    ret i32 0
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l122
// 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 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..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 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !47
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK17-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..19
// CHECK17-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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK17-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK17-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..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 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !53
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK17-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..23
// CHECK17-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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK17-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK17-SAME: (i32 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..26 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[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:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..27 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !59
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK17-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..27
// CHECK17-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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK17-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK17-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK17-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK17-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK17-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..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 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..31 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !65
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP66:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK17-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..31
// CHECK17-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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK17-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK17-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK17-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK17-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP69:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK17-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK17-SAME: (i32 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..34 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[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:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK17:       cond.true:
// CHECK17-NEXT:    br label [[COND_END:%.*]]
// CHECK17:       cond.false:
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    br label [[COND_END]]
// CHECK17:       cond.end:
// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..35 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !71
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP72:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK17:       omp.loop.exit:
// CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK17-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..35
// CHECK17-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]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
// CHECK17-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK17:       omp.dispatch.cond:
// CHECK17-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK17-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK17-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK17:       omp.dispatch.body:
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK17-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK17:       omp.inner.for.cond:
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK17:       omp.inner.for.body:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK17:       omp.body.continue:
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK17:       omp.inner.for.inc:
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK17-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP75:![0-9]+]]
// CHECK17:       omp.inner.for.end:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK17:       omp.dispatch.inc:
// CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK17:       omp.dispatch.end:
// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK17-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK17-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK17:       .omp.final.then:
// CHECK17-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK17:       .omp.final.done:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK17-SAME: () #[[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_l154.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_l154(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_l159.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_l159(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_l164.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_l164(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_l169.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_l169(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_l174.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_l174(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_l154
// 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, !llvm.access.group !14
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !14
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK19-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK19-SAME: (i32* noalias 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, !llvm.access.group !18
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK19-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK19-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK19-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK19-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK19-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l159
// CHECK19-SAME: (i32 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, !llvm.access.group !23
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !23
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK19-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK19-SAME: (i32* noalias 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, !llvm.access.group !26
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK19-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK19-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK19-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK19-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK19-NEXT:    store i32 [[ADD10]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l164
// CHECK19-SAME: (i32 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, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK19-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
// CHECK19-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]]), !llvm.access.group !29
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK19-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK19-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
// CHECK19-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK19:       cond.true11:
// CHECK19-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    br label [[COND_END13:%.*]]
// CHECK19:       cond.false12:
// CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    br label [[COND_END13]]
// CHECK19:       cond.end13:
// CHECK19-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
// CHECK19-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
// CHECK19-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK19-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK19-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
// CHECK19-NEXT:    [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD17]], i32* [[I4]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..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, !llvm.access.group !32
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK19-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK19-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK19-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK19-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
// CHECK19-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK19-NEXT:    store i32 [[ADD11]], i32* [[I4]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l169
// CHECK19-SAME: (i32 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, !llvm.access.group !35
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !35
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK19-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..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 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK19-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK19-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK19-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK19-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK19-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK19-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK19-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK19-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l174
// CHECK19-SAME: (i32 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, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK19-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]]), !llvm.access.group !41
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK19-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK19-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK19-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD9]], i32* [[I4]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..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 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK19-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK19-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK19-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK19-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
// CHECK19-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
// CHECK19-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
// CHECK19-NEXT:    store i32 [[ADD10]], i32* [[I4]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK19:       omp.precond.end:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK19-SAME: (i32 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__l122.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__l122([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__l127.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__l127([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__l132.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__l132(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__l137.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__l137([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__l142.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__l142(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__l122
// 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, !llvm.access.group !47
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !47
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK19-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..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, !llvm.access.group !50
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK19-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l127
// CHECK19-SAME: ([10 x i32]* 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, !llvm.access.group !53
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !53
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK19-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..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, !llvm.access.group !56
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK19-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l132
// CHECK19-SAME: (i32 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, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..27 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !59
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK19-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..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, !llvm.access.group !62
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK19-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
// CHECK19-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK19-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK19-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK19-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l137
// CHECK19-SAME: ([10 x i32]* 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, !llvm.access.group !65
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..31 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !65
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !65
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP66:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK19-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..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 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK19-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
// CHECK19-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
// CHECK19-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !68
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP69:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK19-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l142
// CHECK19-SAME: (i32 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, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..35 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !71
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !71
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP72:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK19:       omp.loop.exit:
// CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK19-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..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 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK19:       omp.dispatch.cond:
// CHECK19-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
// CHECK19-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK19-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK19:       omp.dispatch.body:
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK19-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK19:       omp.inner.for.cond:
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK19:       omp.inner.for.body:
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
// CHECK19-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK19:       omp.body.continue:
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK19:       omp.inner.for.inc:
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK19-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !74
// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP75:![0-9]+]]
// CHECK19:       omp.inner.for.end:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK19:       omp.dispatch.inc:
// CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK19:       omp.dispatch.end:
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK19-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK19:       .omp.final.then:
// CHECK19-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK19:       .omp.final.done:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK19-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK19-NEXT:    ret void
//
//
// CHECK21-LABEL: define {{[^@]+}}@main
// CHECK21-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK21-NEXT:  entry:
// CHECK21-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK21-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK21-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK21-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB16:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB17:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I18:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP40:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB46:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB47:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I48:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV51:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I52:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP69:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_70:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_71:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB75:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB76:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I77:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV80:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I81:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_98:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP99:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_100:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_101:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB105:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB106:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I107:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV110:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I111:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK21-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK21-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK21-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK21-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK21-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK21-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
// CHECK21-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
// CHECK21-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK21-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK21-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
// CHECK21-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK21-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK21-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK21-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK21-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK21-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
// CHECK21-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
// CHECK21:       simd.if.then:
// CHECK21-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK21-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK21:       omp.inner.for.cond:
// CHECK21-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK21-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK21:       omp.inner.for.body:
// CHECK21-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK21-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK21-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK21-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK21:       omp.body.continue:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK21:       omp.inner.for.inc:
// CHECK21-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK21-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK21:       omp.inner.for.end:
// CHECK21-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP13]], 0
// CHECK21-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK21-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK21-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK21-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END]]
// CHECK21:       simd.if.end:
// CHECK21-NEXT:    [[TMP14:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK21-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK21-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[TMP15]], 0
// CHECK21-NEXT:    [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
// CHECK21-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
// CHECK21-NEXT:    store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB16]], align 4
// CHECK21-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK21-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_UB17]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I18]], align 4
// CHECK21-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK21-NEXT:    [[CMP19:%.*]] = icmp slt i32 0, [[TMP17]]
// CHECK21-NEXT:    br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END38:%.*]]
// CHECK21:       simd.if.then20:
// CHECK21-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
// CHECK21-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV21]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK21:       omp.inner.for.cond23:
// CHECK21-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK21-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
// CHECK21:       omp.inner.for.body25:
// CHECK21-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK21-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK21-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[IDXPROM28:%.*]] = sext i32 [[TMP22]] to i64
// CHECK21-NEXT:    [[ARRAYIDX29:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM28]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE30:%.*]]
// CHECK21:       omp.body.continue30:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
// CHECK21:       omp.inner.for.inc31:
// CHECK21-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK21-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK21:       omp.inner.for.end33:
// CHECK21-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK21-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[TMP24]], 0
// CHECK21-NEXT:    [[DIV35:%.*]] = sdiv i32 [[SUB34]], 1
// CHECK21-NEXT:    [[MUL36:%.*]] = mul nsw i32 [[DIV35]], 1
// CHECK21-NEXT:    [[ADD37:%.*]] = add nsw i32 0, [[MUL36]]
// CHECK21-NEXT:    store i32 [[ADD37]], i32* [[I22]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END38]]
// CHECK21:       simd.if.end38:
// CHECK21-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_39]], align 4
// CHECK21-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK21-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK21-NEXT:    [[SUB43:%.*]] = sub nsw i32 [[TMP27]], 0
// CHECK21-NEXT:    [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
// CHECK21-NEXT:    [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
// CHECK21-NEXT:    store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB46]], align 4
// CHECK21-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
// CHECK21-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_UB47]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I48]], align 4
// CHECK21-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK21-NEXT:    [[CMP49:%.*]] = icmp slt i32 0, [[TMP29]]
// CHECK21-NEXT:    br i1 [[CMP49]], label [[SIMD_IF_THEN50:%.*]], label [[SIMD_IF_END68:%.*]]
// CHECK21:       simd.if.then50:
// CHECK21-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_LB46]], align 4
// CHECK21-NEXT:    store i32 [[TMP30]], i32* [[DOTOMP_IV51]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND53:%.*]]
// CHECK21:       omp.inner.for.cond53:
// CHECK21-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_UB47]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[CMP54:%.*]] = icmp sle i32 [[TMP31]], [[TMP32]]
// CHECK21-NEXT:    br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END63:%.*]]
// CHECK21:       omp.inner.for.body55:
// CHECK21-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[MUL56:%.*]] = mul nsw i32 [[TMP33]], 1
// CHECK21-NEXT:    [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
// CHECK21-NEXT:    store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[TMP34:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[IDXPROM58:%.*]] = sext i32 [[TMP34]] to i64
// CHECK21-NEXT:    [[ARRAYIDX59:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM58]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX59]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE60:%.*]]
// CHECK21:       omp.body.continue60:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC61:%.*]]
// CHECK21:       omp.inner.for.inc61:
// CHECK21-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    [[ADD62:%.*]] = add nsw i32 [[TMP35]], 1
// CHECK21-NEXT:    store i32 [[ADD62]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK21:       omp.inner.for.end63:
// CHECK21-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK21-NEXT:    [[SUB64:%.*]] = sub nsw i32 [[TMP36]], 0
// CHECK21-NEXT:    [[DIV65:%.*]] = sdiv i32 [[SUB64]], 1
// CHECK21-NEXT:    [[MUL66:%.*]] = mul nsw i32 [[DIV65]], 1
// CHECK21-NEXT:    [[ADD67:%.*]] = add nsw i32 0, [[MUL66]]
// CHECK21-NEXT:    store i32 [[ADD67]], i32* [[I52]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END68]]
// CHECK21:       simd.if.end68:
// CHECK21-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK21-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK21-NEXT:    [[SUB72:%.*]] = sub nsw i32 [[TMP38]], 0
// CHECK21-NEXT:    [[DIV73:%.*]] = sdiv i32 [[SUB72]], 1
// CHECK21-NEXT:    [[SUB74:%.*]] = sub nsw i32 [[DIV73]], 1
// CHECK21-NEXT:    store i32 [[SUB74]], i32* [[DOTCAPTURE_EXPR_71]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB75]], align 4
// CHECK21-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_71]], align 4
// CHECK21-NEXT:    store i32 [[TMP39]], i32* [[DOTOMP_UB76]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I77]], align 4
// CHECK21-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK21-NEXT:    [[CMP78:%.*]] = icmp slt i32 0, [[TMP40]]
// CHECK21-NEXT:    br i1 [[CMP78]], label [[SIMD_IF_THEN79:%.*]], label [[SIMD_IF_END97:%.*]]
// CHECK21:       simd.if.then79:
// CHECK21-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_LB75]], align 4
// CHECK21-NEXT:    store i32 [[TMP41]], i32* [[DOTOMP_IV80]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND82:%.*]]
// CHECK21:       omp.inner.for.cond82:
// CHECK21-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_UB76]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[CMP83:%.*]] = icmp sle i32 [[TMP42]], [[TMP43]]
// CHECK21-NEXT:    br i1 [[CMP83]], label [[OMP_INNER_FOR_BODY84:%.*]], label [[OMP_INNER_FOR_END92:%.*]]
// CHECK21:       omp.inner.for.body84:
// CHECK21-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[MUL85:%.*]] = mul nsw i32 [[TMP44]], 1
// CHECK21-NEXT:    [[ADD86:%.*]] = add nsw i32 0, [[MUL85]]
// CHECK21-NEXT:    store i32 [[ADD86]], i32* [[I81]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[TMP45:%.*]] = load i32, i32* [[I81]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[IDXPROM87:%.*]] = sext i32 [[TMP45]] to i64
// CHECK21-NEXT:    [[ARRAYIDX88:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM87]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX88]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE89:%.*]]
// CHECK21:       omp.body.continue89:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC90:%.*]]
// CHECK21:       omp.inner.for.inc90:
// CHECK21-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    [[ADD91:%.*]] = add nsw i32 [[TMP46]], 1
// CHECK21-NEXT:    store i32 [[ADD91]], i32* [[DOTOMP_IV80]], align 4, !llvm.access.group !12
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND82]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK21:       omp.inner.for.end92:
// CHECK21-NEXT:    [[TMP47:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_70]], align 4
// CHECK21-NEXT:    [[SUB93:%.*]] = sub nsw i32 [[TMP47]], 0
// CHECK21-NEXT:    [[DIV94:%.*]] = sdiv i32 [[SUB93]], 1
// CHECK21-NEXT:    [[MUL95:%.*]] = mul nsw i32 [[DIV94]], 1
// CHECK21-NEXT:    [[ADD96:%.*]] = add nsw i32 0, [[MUL95]]
// CHECK21-NEXT:    store i32 [[ADD96]], i32* [[I81]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END97]]
// CHECK21:       simd.if.end97:
// CHECK21-NEXT:    [[TMP48:%.*]] = load i32, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 [[TMP48]], i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK21-NEXT:    [[TMP49:%.*]] = load i32, i32* [[N]], align 4
// CHECK21-NEXT:    store i32 [[TMP49]], i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK21-NEXT:    [[TMP50:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK21-NEXT:    [[SUB102:%.*]] = sub nsw i32 [[TMP50]], 0
// CHECK21-NEXT:    [[DIV103:%.*]] = sdiv i32 [[SUB102]], 1
// CHECK21-NEXT:    [[SUB104:%.*]] = sub nsw i32 [[DIV103]], 1
// CHECK21-NEXT:    store i32 [[SUB104]], i32* [[DOTCAPTURE_EXPR_101]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB105]], align 4
// CHECK21-NEXT:    [[TMP51:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_101]], align 4
// CHECK21-NEXT:    store i32 [[TMP51]], i32* [[DOTOMP_UB106]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[I107]], align 4
// CHECK21-NEXT:    [[TMP52:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK21-NEXT:    [[CMP108:%.*]] = icmp slt i32 0, [[TMP52]]
// CHECK21-NEXT:    br i1 [[CMP108]], label [[SIMD_IF_THEN109:%.*]], label [[SIMD_IF_END127:%.*]]
// CHECK21:       simd.if.then109:
// CHECK21-NEXT:    [[TMP53:%.*]] = load i32, i32* [[DOTOMP_LB105]], align 4
// CHECK21-NEXT:    store i32 [[TMP53]], i32* [[DOTOMP_IV110]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND112:%.*]]
// CHECK21:       omp.inner.for.cond112:
// CHECK21-NEXT:    [[TMP54:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[TMP55:%.*]] = load i32, i32* [[DOTOMP_UB106]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[CMP113:%.*]] = icmp sle i32 [[TMP54]], [[TMP55]]
// CHECK21-NEXT:    br i1 [[CMP113]], label [[OMP_INNER_FOR_BODY114:%.*]], label [[OMP_INNER_FOR_END122:%.*]]
// CHECK21:       omp.inner.for.body114:
// CHECK21-NEXT:    [[TMP56:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[MUL115:%.*]] = mul nsw i32 [[TMP56]], 1
// CHECK21-NEXT:    [[ADD116:%.*]] = add nsw i32 0, [[MUL115]]
// CHECK21-NEXT:    store i32 [[ADD116]], i32* [[I111]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[TMP57:%.*]] = load i32, i32* [[I111]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[IDXPROM117:%.*]] = sext i32 [[TMP57]] to i64
// CHECK21-NEXT:    [[ARRAYIDX118:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM117]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX118]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE119:%.*]]
// CHECK21:       omp.body.continue119:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC120:%.*]]
// CHECK21:       omp.inner.for.inc120:
// CHECK21-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    [[ADD121:%.*]] = add nsw i32 [[TMP58]], 1
// CHECK21-NEXT:    store i32 [[ADD121]], i32* [[DOTOMP_IV110]], align 4, !llvm.access.group !15
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND112]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK21:       omp.inner.for.end122:
// CHECK21-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_100]], align 4
// CHECK21-NEXT:    [[SUB123:%.*]] = sub nsw i32 [[TMP59]], 0
// CHECK21-NEXT:    [[DIV124:%.*]] = sdiv i32 [[SUB123]], 1
// CHECK21-NEXT:    [[MUL125:%.*]] = mul nsw i32 [[DIV124]], 1
// CHECK21-NEXT:    [[ADD126:%.*]] = add nsw i32 0, [[MUL125]]
// CHECK21-NEXT:    store i32 [[ADD126]], i32* [[I111]], align 4
// CHECK21-NEXT:    br label [[SIMD_IF_END127]]
// CHECK21:       simd.if.end127:
// CHECK21-NEXT:    [[TMP60:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK21-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP60]])
// CHECK21-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK21-NEXT:    [[TMP61:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK21-NEXT:    call void @llvm.stackrestore(i8* [[TMP61]])
// CHECK21-NEXT:    [[TMP62:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK21-NEXT:    ret i32 [[TMP62]]
//
//
// CHECK21-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK21-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
// CHECK21-NEXT:  entry:
// CHECK21-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK21-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB19:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB20:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP34:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB35:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB36:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV37:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I38:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTCAPTURE_EXPR_50:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[_TMP51:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_LB52:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_UB53:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[DOTOMP_IV54:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    [[I55:%.*]] = alloca i32, align 4
// CHECK21-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK21-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK21-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK21-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK21:       omp.inner.for.cond:
// CHECK21-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK21-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK21:       omp.inner.for.body:
// CHECK21-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK21-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK21-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
// CHECK21-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK21:       omp.body.continue:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK21:       omp.inner.for.inc:
// CHECK21-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK21-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK21:       omp.inner.for.end:
// CHECK21-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB4]], align 4
// CHECK21-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
// CHECK21-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK21:       omp.inner.for.cond7:
// CHECK21-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK21-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
// CHECK21:       omp.inner.for.body9:
// CHECK21-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK21-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK21-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP10]] to i64
// CHECK21-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM12]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX13]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE14:%.*]]
// CHECK21:       omp.body.continue14:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC15:%.*]]
// CHECK21:       omp.inner.for.inc15:
// CHECK21-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK21-NEXT:    store i32 [[ADD16]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK21:       omp.inner.for.end17:
// CHECK21-NEXT:    store i32 10, i32* [[I6]], align 4
// CHECK21-NEXT:    [[TMP12:%.*]] = load i32, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB19]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB20]], align 4
// CHECK21-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB19]], align 4
// CHECK21-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV21]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK21:       omp.inner.for.cond23:
// CHECK21-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB20]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK21-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
// CHECK21:       omp.inner.for.body25:
// CHECK21-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK21-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK21-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[IDXPROM28:%.*]] = sext i32 [[TMP17]] to i64
// CHECK21-NEXT:    [[ARRAYIDX29:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM28]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE30:%.*]]
// CHECK21:       omp.body.continue30:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
// CHECK21:       omp.inner.for.inc31:
// CHECK21-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK21-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !24
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK21:       omp.inner.for.end33:
// CHECK21-NEXT:    store i32 10, i32* [[I22]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB35]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB36]], align 4
// CHECK21-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB35]], align 4
// CHECK21-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV37]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND39:%.*]]
// CHECK21:       omp.inner.for.cond39:
// CHECK21-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB36]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[CMP40:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK21-NEXT:    br i1 [[CMP40]], label [[OMP_INNER_FOR_BODY41:%.*]], label [[OMP_INNER_FOR_END49:%.*]]
// CHECK21:       omp.inner.for.body41:
// CHECK21-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[MUL42:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK21-NEXT:    [[ADD43:%.*]] = add nsw i32 0, [[MUL42]]
// CHECK21-NEXT:    store i32 [[ADD43]], i32* [[I38]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I38]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[IDXPROM44:%.*]] = sext i32 [[TMP23]] to i64
// CHECK21-NEXT:    [[ARRAYIDX45:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM44]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX45]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE46:%.*]]
// CHECK21:       omp.body.continue46:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC47:%.*]]
// CHECK21:       omp.inner.for.inc47:
// CHECK21-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    [[ADD48:%.*]] = add nsw i32 [[TMP24]], 1
// CHECK21-NEXT:    store i32 [[ADD48]], i32* [[DOTOMP_IV37]], align 4, !llvm.access.group !27
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND39]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK21:       omp.inner.for.end49:
// CHECK21-NEXT:    store i32 10, i32* [[I38]], align 4
// CHECK21-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK21-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_50]], align 4
// CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB52]], align 4
// CHECK21-NEXT:    store i32 9, i32* [[DOTOMP_UB53]], align 4
// CHECK21-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB52]], align 4
// CHECK21-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV54]], align 4
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND56:%.*]]
// CHECK21:       omp.inner.for.cond56:
// CHECK21-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB53]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[CMP57:%.*]] = icmp sle i32 [[TMP27]], [[TMP28]]
// CHECK21-NEXT:    br i1 [[CMP57]], label [[OMP_INNER_FOR_BODY58:%.*]], label [[OMP_INNER_FOR_END66:%.*]]
// CHECK21:       omp.inner.for.body58:
// CHECK21-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[MUL59:%.*]] = mul nsw i32 [[TMP29]], 1
// CHECK21-NEXT:    [[ADD60:%.*]] = add nsw i32 0, [[MUL59]]
// CHECK21-NEXT:    store i32 [[ADD60]], i32* [[I55]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I55]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[IDXPROM61:%.*]] = sext i32 [[TMP30]] to i64
// CHECK21-NEXT:    [[ARRAYIDX62:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM61]]
// CHECK21-NEXT:    store i32 0, i32* [[ARRAYIDX62]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE63:%.*]]
// CHECK21:       omp.body.continue63:
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC64:%.*]]
// CHECK21:       omp.inner.for.inc64:
// CHECK21-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    [[ADD65:%.*]] = add nsw i32 [[TMP31]], 1
// CHECK21-NEXT:    store i32 [[ADD65]], i32* [[DOTOMP_IV54]], align 4, !llvm.access.group !30
// CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND56]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK21:       omp.inner.for.end66:
// CHECK21-NEXT:    store i32 10, i32* [[I55]], align 4
// CHECK21-NEXT:    ret i32 0
//
//
// CHECK23-LABEL: define {{[^@]+}}@main
// CHECK23-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK23-NEXT:  entry:
// CHECK23-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 4
// CHECK23-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK23-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB16:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB17:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I18:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV21:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I22:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP39:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB45:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB46:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I47:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV50:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I51:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP67:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_68:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_69:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB73:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB74:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I75:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV78:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I79:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_95:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP96:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_97:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_98:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB102:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB103:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I104:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV107:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I108:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK23-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK23-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK23-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK23-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
// CHECK23-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
// CHECK23-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
// CHECK23-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK23-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
// CHECK23-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK23-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK23-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK23-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
// CHECK23-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
// CHECK23:       simd.if.then:
// CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK23-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK23:       omp.inner.for.cond:
// CHECK23-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK23-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK23:       omp.inner.for.body:
// CHECK23-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK23-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK23-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP10]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK23:       omp.body.continue:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK23:       omp.inner.for.inc:
// CHECK23-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK23-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK23:       omp.inner.for.end:
// CHECK23-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP12]], 0
// CHECK23-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK23-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
// CHECK23-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK23-NEXT:    store i32 [[ADD9]], i32* [[I3]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END]]
// CHECK23:       simd.if.end:
// CHECK23-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP13]], i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK23-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK23-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[TMP14]], 0
// CHECK23-NEXT:    [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
// CHECK23-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
// CHECK23-NEXT:    store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB16]], align 4
// CHECK23-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK23-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_UB17]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I18]], align 4
// CHECK23-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK23-NEXT:    [[CMP19:%.*]] = icmp slt i32 0, [[TMP16]]
// CHECK23-NEXT:    br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END37:%.*]]
// CHECK23:       simd.if.then20:
// CHECK23-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
// CHECK23-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV21]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND23:%.*]]
// CHECK23:       omp.inner.for.cond23:
// CHECK23-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[CMP24:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK23-NEXT:    br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END32:%.*]]
// CHECK23:       omp.inner.for.body25:
// CHECK23-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[TMP20]], 1
// CHECK23-NEXT:    [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
// CHECK23-NEXT:    store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP21]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX28]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE29:%.*]]
// CHECK23:       omp.body.continue29:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC30:%.*]]
// CHECK23:       omp.inner.for.inc30:
// CHECK23-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK23-NEXT:    store i32 [[ADD31]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK23:       omp.inner.for.end32:
// CHECK23-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
// CHECK23-NEXT:    [[SUB33:%.*]] = sub nsw i32 [[TMP23]], 0
// CHECK23-NEXT:    [[DIV34:%.*]] = sdiv i32 [[SUB33]], 1
// CHECK23-NEXT:    [[MUL35:%.*]] = mul nsw i32 [[DIV34]], 1
// CHECK23-NEXT:    [[ADD36:%.*]] = add nsw i32 0, [[MUL35]]
// CHECK23-NEXT:    store i32 [[ADD36]], i32* [[I22]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END37]]
// CHECK23:       simd.if.end37:
// CHECK23-NEXT:    [[TMP24:%.*]] = load i32, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 [[TMP24]], i32* [[DOTCAPTURE_EXPR_38]], align 4
// CHECK23-NEXT:    [[TMP25:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK23-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK23-NEXT:    [[SUB42:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK23-NEXT:    [[DIV43:%.*]] = sdiv i32 [[SUB42]], 1
// CHECK23-NEXT:    [[SUB44:%.*]] = sub nsw i32 [[DIV43]], 1
// CHECK23-NEXT:    store i32 [[SUB44]], i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB45]], align 4
// CHECK23-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
// CHECK23-NEXT:    store i32 [[TMP27]], i32* [[DOTOMP_UB46]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I47]], align 4
// CHECK23-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK23-NEXT:    [[CMP48:%.*]] = icmp slt i32 0, [[TMP28]]
// CHECK23-NEXT:    br i1 [[CMP48]], label [[SIMD_IF_THEN49:%.*]], label [[SIMD_IF_END66:%.*]]
// CHECK23:       simd.if.then49:
// CHECK23-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_LB45]], align 4
// CHECK23-NEXT:    store i32 [[TMP29]], i32* [[DOTOMP_IV50]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND52:%.*]]
// CHECK23:       omp.inner.for.cond52:
// CHECK23-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_UB46]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[CMP53:%.*]] = icmp sle i32 [[TMP30]], [[TMP31]]
// CHECK23-NEXT:    br i1 [[CMP53]], label [[OMP_INNER_FOR_BODY54:%.*]], label [[OMP_INNER_FOR_END61:%.*]]
// CHECK23:       omp.inner.for.body54:
// CHECK23-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[MUL55:%.*]] = mul nsw i32 [[TMP32]], 1
// CHECK23-NEXT:    [[ADD56:%.*]] = add nsw i32 0, [[MUL55]]
// CHECK23-NEXT:    store i32 [[ADD56]], i32* [[I51]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[TMP33:%.*]] = load i32, i32* [[I51]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[ARRAYIDX57:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP33]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX57]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE58:%.*]]
// CHECK23:       omp.body.continue58:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC59:%.*]]
// CHECK23:       omp.inner.for.inc59:
// CHECK23-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP34]], 1
// CHECK23-NEXT:    store i32 [[ADD60]], i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND52]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK23:       omp.inner.for.end61:
// CHECK23-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
// CHECK23-NEXT:    [[SUB62:%.*]] = sub nsw i32 [[TMP35]], 0
// CHECK23-NEXT:    [[DIV63:%.*]] = sdiv i32 [[SUB62]], 1
// CHECK23-NEXT:    [[MUL64:%.*]] = mul nsw i32 [[DIV63]], 1
// CHECK23-NEXT:    [[ADD65:%.*]] = add nsw i32 0, [[MUL64]]
// CHECK23-NEXT:    store i32 [[ADD65]], i32* [[I51]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END66]]
// CHECK23:       simd.if.end66:
// CHECK23-NEXT:    [[TMP36:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK23-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK23-NEXT:    [[SUB70:%.*]] = sub nsw i32 [[TMP37]], 0
// CHECK23-NEXT:    [[DIV71:%.*]] = sdiv i32 [[SUB70]], 1
// CHECK23-NEXT:    [[SUB72:%.*]] = sub nsw i32 [[DIV71]], 1
// CHECK23-NEXT:    store i32 [[SUB72]], i32* [[DOTCAPTURE_EXPR_69]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB73]], align 4
// CHECK23-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_69]], align 4
// CHECK23-NEXT:    store i32 [[TMP38]], i32* [[DOTOMP_UB74]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I75]], align 4
// CHECK23-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK23-NEXT:    [[CMP76:%.*]] = icmp slt i32 0, [[TMP39]]
// CHECK23-NEXT:    br i1 [[CMP76]], label [[SIMD_IF_THEN77:%.*]], label [[SIMD_IF_END94:%.*]]
// CHECK23:       simd.if.then77:
// CHECK23-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_LB73]], align 4
// CHECK23-NEXT:    store i32 [[TMP40]], i32* [[DOTOMP_IV78]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND80:%.*]]
// CHECK23:       omp.inner.for.cond80:
// CHECK23-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_UB74]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[CMP81:%.*]] = icmp sle i32 [[TMP41]], [[TMP42]]
// CHECK23-NEXT:    br i1 [[CMP81]], label [[OMP_INNER_FOR_BODY82:%.*]], label [[OMP_INNER_FOR_END89:%.*]]
// CHECK23:       omp.inner.for.body82:
// CHECK23-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[MUL83:%.*]] = mul nsw i32 [[TMP43]], 1
// CHECK23-NEXT:    [[ADD84:%.*]] = add nsw i32 0, [[MUL83]]
// CHECK23-NEXT:    store i32 [[ADD84]], i32* [[I79]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[TMP44:%.*]] = load i32, i32* [[I79]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[ARRAYIDX85:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP44]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX85]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE86:%.*]]
// CHECK23:       omp.body.continue86:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC87:%.*]]
// CHECK23:       omp.inner.for.inc87:
// CHECK23-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    [[ADD88:%.*]] = add nsw i32 [[TMP45]], 1
// CHECK23-NEXT:    store i32 [[ADD88]], i32* [[DOTOMP_IV78]], align 4, !llvm.access.group !13
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND80]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK23:       omp.inner.for.end89:
// CHECK23-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_68]], align 4
// CHECK23-NEXT:    [[SUB90:%.*]] = sub nsw i32 [[TMP46]], 0
// CHECK23-NEXT:    [[DIV91:%.*]] = sdiv i32 [[SUB90]], 1
// CHECK23-NEXT:    [[MUL92:%.*]] = mul nsw i32 [[DIV91]], 1
// CHECK23-NEXT:    [[ADD93:%.*]] = add nsw i32 0, [[MUL92]]
// CHECK23-NEXT:    store i32 [[ADD93]], i32* [[I79]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END94]]
// CHECK23:       simd.if.end94:
// CHECK23-NEXT:    [[TMP47:%.*]] = load i32, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 [[TMP47]], i32* [[DOTCAPTURE_EXPR_95]], align 4
// CHECK23-NEXT:    [[TMP48:%.*]] = load i32, i32* [[N]], align 4
// CHECK23-NEXT:    store i32 [[TMP48]], i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK23-NEXT:    [[TMP49:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK23-NEXT:    [[SUB99:%.*]] = sub nsw i32 [[TMP49]], 0
// CHECK23-NEXT:    [[DIV100:%.*]] = sdiv i32 [[SUB99]], 1
// CHECK23-NEXT:    [[SUB101:%.*]] = sub nsw i32 [[DIV100]], 1
// CHECK23-NEXT:    store i32 [[SUB101]], i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB102]], align 4
// CHECK23-NEXT:    [[TMP50:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_98]], align 4
// CHECK23-NEXT:    store i32 [[TMP50]], i32* [[DOTOMP_UB103]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[I104]], align 4
// CHECK23-NEXT:    [[TMP51:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK23-NEXT:    [[CMP105:%.*]] = icmp slt i32 0, [[TMP51]]
// CHECK23-NEXT:    br i1 [[CMP105]], label [[SIMD_IF_THEN106:%.*]], label [[SIMD_IF_END123:%.*]]
// CHECK23:       simd.if.then106:
// CHECK23-NEXT:    [[TMP52:%.*]] = load i32, i32* [[DOTOMP_LB102]], align 4
// CHECK23-NEXT:    store i32 [[TMP52]], i32* [[DOTOMP_IV107]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND109:%.*]]
// CHECK23:       omp.inner.for.cond109:
// CHECK23-NEXT:    [[TMP53:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[TMP54:%.*]] = load i32, i32* [[DOTOMP_UB103]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[CMP110:%.*]] = icmp sle i32 [[TMP53]], [[TMP54]]
// CHECK23-NEXT:    br i1 [[CMP110]], label [[OMP_INNER_FOR_BODY111:%.*]], label [[OMP_INNER_FOR_END118:%.*]]
// CHECK23:       omp.inner.for.body111:
// CHECK23-NEXT:    [[TMP55:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[MUL112:%.*]] = mul nsw i32 [[TMP55]], 1
// CHECK23-NEXT:    [[ADD113:%.*]] = add nsw i32 0, [[MUL112]]
// CHECK23-NEXT:    store i32 [[ADD113]], i32* [[I108]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[TMP56:%.*]] = load i32, i32* [[I108]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[ARRAYIDX114:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP56]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX114]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE115:%.*]]
// CHECK23:       omp.body.continue115:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC116:%.*]]
// CHECK23:       omp.inner.for.inc116:
// CHECK23-NEXT:    [[TMP57:%.*]] = load i32, i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    [[ADD117:%.*]] = add nsw i32 [[TMP57]], 1
// CHECK23-NEXT:    store i32 [[ADD117]], i32* [[DOTOMP_IV107]], align 4, !llvm.access.group !16
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND109]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK23:       omp.inner.for.end118:
// CHECK23-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_97]], align 4
// CHECK23-NEXT:    [[SUB119:%.*]] = sub nsw i32 [[TMP58]], 0
// CHECK23-NEXT:    [[DIV120:%.*]] = sdiv i32 [[SUB119]], 1
// CHECK23-NEXT:    [[MUL121:%.*]] = mul nsw i32 [[DIV120]], 1
// CHECK23-NEXT:    [[ADD122:%.*]] = add nsw i32 0, [[MUL121]]
// CHECK23-NEXT:    store i32 [[ADD122]], i32* [[I108]], align 4
// CHECK23-NEXT:    br label [[SIMD_IF_END123]]
// CHECK23:       simd.if.end123:
// CHECK23-NEXT:    [[TMP59:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK23-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP59]])
// CHECK23-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK23-NEXT:    [[TMP60:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK23-NEXT:    call void @llvm.stackrestore(i8* [[TMP60]])
// CHECK23-NEXT:    [[TMP61:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK23-NEXT:    ret i32 [[TMP61]]
//
//
// CHECK23-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
// CHECK23-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
// CHECK23-NEXT:  entry:
// CHECK23-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK23-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP17:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB18:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB19:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV20:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I21:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB33:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB34:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV35:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I36:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTCAPTURE_EXPR_47:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[_TMP48:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_LB49:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_UB50:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[DOTOMP_IV51:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    [[I52:%.*]] = alloca i32, align 4
// CHECK23-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK23-NEXT:    store i32 10, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK23-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK23-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK23:       omp.inner.for.cond:
// CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK23-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK23:       omp.inner.for.body:
// CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK23-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK23-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP4]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK23:       omp.body.continue:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK23:       omp.inner.for.inc:
// CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
// CHECK23-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
// CHECK23:       omp.inner.for.end:
// CHECK23-NEXT:    store i32 10, i32* [[I]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB4]], align 4
// CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
// CHECK23-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK23:       omp.inner.for.cond7:
// CHECK23-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK23-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
// CHECK23:       omp.inner.for.body9:
// CHECK23-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK23-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK23-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP10]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX12]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE13:%.*]]
// CHECK23:       omp.body.continue13:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC14:%.*]]
// CHECK23:       omp.inner.for.inc14:
// CHECK23-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK23-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !22
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK23:       omp.inner.for.end16:
// CHECK23-NEXT:    store i32 10, i32* [[I6]], align 4
// CHECK23-NEXT:    [[TMP12:%.*]] = load i32, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB18]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB19]], align 4
// CHECK23-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB18]], align 4
// CHECK23-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV20]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
// CHECK23:       omp.inner.for.cond22:
// CHECK23-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB19]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK23-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END31:%.*]]
// CHECK23:       omp.inner.for.body24:
// CHECK23-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK23-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
// CHECK23-NEXT:    store i32 [[ADD26]], i32* [[I21]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I21]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[ARRAYIDX27:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP17]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX27]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE28:%.*]]
// CHECK23:       omp.body.continue28:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC29:%.*]]
// CHECK23:       omp.inner.for.inc29:
// CHECK23-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    [[ADD30:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK23-NEXT:    store i32 [[ADD30]], i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !25
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK23:       omp.inner.for.end31:
// CHECK23-NEXT:    store i32 10, i32* [[I21]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB33]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB34]], align 4
// CHECK23-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB33]], align 4
// CHECK23-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV35]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND37:%.*]]
// CHECK23:       omp.inner.for.cond37:
// CHECK23-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB34]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK23-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY39:%.*]], label [[OMP_INNER_FOR_END46:%.*]]
// CHECK23:       omp.inner.for.body39:
// CHECK23-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[MUL40:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK23-NEXT:    [[ADD41:%.*]] = add nsw i32 0, [[MUL40]]
// CHECK23-NEXT:    store i32 [[ADD41]], i32* [[I36]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I36]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[ARRAYIDX42:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP23]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX42]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE43:%.*]]
// CHECK23:       omp.body.continue43:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC44:%.*]]
// CHECK23:       omp.inner.for.inc44:
// CHECK23-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    [[ADD45:%.*]] = add nsw i32 [[TMP24]], 1
// CHECK23-NEXT:    store i32 [[ADD45]], i32* [[DOTOMP_IV35]], align 4, !llvm.access.group !28
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND37]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK23:       omp.inner.for.end46:
// CHECK23-NEXT:    store i32 10, i32* [[I36]], align 4
// CHECK23-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M]], align 4
// CHECK23-NEXT:    store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_47]], align 4
// CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB49]], align 4
// CHECK23-NEXT:    store i32 9, i32* [[DOTOMP_UB50]], align 4
// CHECK23-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB49]], align 4
// CHECK23-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV51]], align 4
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND53:%.*]]
// CHECK23:       omp.inner.for.cond53:
// CHECK23-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB50]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[CMP54:%.*]] = icmp sle i32 [[TMP27]], [[TMP28]]
// CHECK23-NEXT:    br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END62:%.*]]
// CHECK23:       omp.inner.for.body55:
// CHECK23-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[MUL56:%.*]] = mul nsw i32 [[TMP29]], 1
// CHECK23-NEXT:    [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
// CHECK23-NEXT:    store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[ARRAYIDX58:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP30]]
// CHECK23-NEXT:    store i32 0, i32* [[ARRAYIDX58]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE59:%.*]]
// CHECK23:       omp.body.continue59:
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC60:%.*]]
// CHECK23:       omp.inner.for.inc60:
// CHECK23-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    [[ADD61:%.*]] = add nsw i32 [[TMP31]], 1
// CHECK23-NEXT:    store i32 [[ADD61]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !31
// CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK23:       omp.inner.for.end62:
// CHECK23-NEXT:    store i32 10, i32* [[I52]], align 4
// CHECK23-NEXT:    ret i32 0
//