nuudlman/llvm: clang/test/OpenMP/target_teams_distribute_parallel_for_collapse

// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
// expected-no-diagnostics
#ifndef HEADER
#define HEADER

// Test host codegen.
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3

// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
#ifdef CK1

template <typename T, int X, long long Y>
struct SS{
  T a[X][Y];

  int foo(void) {

    #pragma omp target teams distribute parallel for collapse(2)
    for(int i = 0; i < X; i++) {
      for(int j = 0; j < Y; j++) {
        a[i][j] = (T)0;
      }
    }

    // discard loop variables not needed here


    return a[0][0];
  }
};

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

}
#endif // CK1

// Test host codegen.
// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
// 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=CHECK9
// 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=CHECK11
// 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=CHECK11

// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
#ifdef CK2

template <typename T, int n, int m>
int tmain(T argc) {
  T a[n][m];
  #pragma omp target teams distribute parallel for collapse(2)
  for(int i = 0; i < n; i++) {
    for(int j = 0; j < m; j++) {
      a[i][j] = (T)0;
    }
  }
  return 0;
}

int main (int argc, char **argv) {
  int n = 100;
  int m = 2;
  int a[n][m];
  #pragma omp target teams distribute parallel for collapse(2)
  for(int i = 0; i < n; i++) {
    for(int j = 0; j < m; j++) {
      a[i][j] = 0;
    }
  }
  return tmain<int, 10, 2>(argc);
}








// discard loop variables not needed here


#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(224352) [[V]])
// CHECK1-NEXT:    ret i32 [[CALL]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK1-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[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:    [[_TMP2:%.*]] = 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 [456 x i32]]**
// CHECK1-NEXT:    store [123 x [456 x i32]]* [[A]], [123 x [456 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 56088, 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_l28.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_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A3]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
// CHECK1-NEXT:    ret i32 [[TMP18]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
// 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:    [[_TMP1:%.*]] = 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:    [[J:%.*]] = 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 56087, 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]], 56087
// 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 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP1:%.*]] = 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:    [[J:%.*]] = 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 56087, 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:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV2]], 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]], 56087
// 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 [ 56087, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 456
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP12]], 456
// CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 456
// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL5]]
// CHECK1-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 0, [[MUL6]]
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[J]], align 4
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[J]], align 4
// CHECK1-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP14]] to i64
// CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i64 0, i64 [[IDXPROM8]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX9]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK1-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
// CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[V]])
// CHECK3-NEXT:    ret i32 [[CALL]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
// CHECK3-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
// CHECK3-NEXT:    store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x [456 x i32]]**
// CHECK3-NEXT:    store [123 x [456 x i32]]* [[A]], [123 x [456 x i32]]** [[TMP3]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK3-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP13]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP14]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK3-NEXT:    store i64 56088, i64* [[TMP15]], align 8
// CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK3-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK3-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A3]], i32 0, i32 0
// CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
// CHECK3-NEXT:    ret i32 [[TMP18]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
// CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 56087, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 56087, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 56087
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 456
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP12]], 456
// CHECK3-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL4]]
// CHECK3-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
// CHECK3-NEXT:    store i32 [[ADD6]], i32* [[J]], align 4
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A]], i32 0, i32 [[TMP13]]
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[J]], align 4
// CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i32 0, i32 [[TMP14]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX7]], align 4
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
// CHECK9-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK9-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK9-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
// CHECK9-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK9-NEXT:    store i32 2, i32* [[M]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[M]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
// CHECK9-NEXT:    [[TMP4:%.*]] = call i8* @llvm.stacksave()
// CHECK9-NEXT:    store i8* [[TMP4]], i8** [[SAVED_STACK]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
// CHECK9-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP5]], align 4
// CHECK9-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
// CHECK9-NEXT:    store i64 [[TMP3]], i64* [[__VLA_EXPR1]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N]], align 4
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP6]], i32* [[CONV]], align 4
// CHECK9-NEXT:    [[TMP7:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[M]], align 4
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP8]], i32* [[CONV1]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
// CHECK9-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
// CHECK9-NEXT:    [[TMP12:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK9-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP12]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes to i8*), i64 40, i1 false)
// CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
// CHECK9-NEXT:    store i64 [[TMP7]], i64* [[TMP14]], align 8
// CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
// CHECK9-NEXT:    store i64 [[TMP7]], i64* [[TMP16]], align 8
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK9-NEXT:    store i8* null, i8** [[TMP17]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK9-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK9-NEXT:    store i64 [[TMP9]], i64* [[TMP19]], align 8
// CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK9-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
// CHECK9-NEXT:    store i64 [[TMP9]], i64* [[TMP21]], align 8
// CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK9-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK9-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
// CHECK9-NEXT:    store i64 [[TMP1]], i64* [[TMP24]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK9-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
// CHECK9-NEXT:    store i64 [[TMP1]], i64* [[TMP26]], align 8
// CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK9-NEXT:    store i8* null, i8** [[TMP27]], align 8
// CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK9-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
// CHECK9-NEXT:    store i64 [[TMP3]], i64* [[TMP29]], align 8
// CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK9-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK9-NEXT:    store i64 [[TMP3]], i64* [[TMP31]], align 8
// CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK9-NEXT:    store i8* null, i8** [[TMP32]], align 8
// CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK9-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i32**
// CHECK9-NEXT:    store i32* [[VLA]], i32** [[TMP34]], align 8
// CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK9-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i32**
// CHECK9-NEXT:    store i32* [[VLA]], i32** [[TMP36]], align 8
// CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK9-NEXT:    store i64 [[TMP11]], i64* [[TMP37]], align 8
// CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK9-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK9-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP42:%.*]] = load i32, i32* [[N]], align 4
// CHECK9-NEXT:    store i32 [[TMP42]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[TMP43:%.*]] = load i32, i32* [[M]], align 4
// CHECK9-NEXT:    store i32 [[TMP43]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK9-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP44]], 0
// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK9-NEXT:    [[CONV5:%.*]] = sext i32 [[DIV]] to i64
// CHECK9-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP45]], 0
// CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK9-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV5]], [[CONV8]]
// CHECK9-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
// CHECK9-NEXT:    store i64 [[SUB9]], i64* [[DOTCAPTURE_EXPR_4]], align 8
// CHECK9-NEXT:    [[TMP46:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_4]], align 8
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP46]], 1
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 1, i32* [[TMP47]], align 4
// CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 5, i32* [[TMP48]], align 4
// CHECK9-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store i8** [[TMP39]], i8*** [[TMP49]], align 8
// CHECK9-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store i8** [[TMP40]], i8*** [[TMP50]], align 8
// CHECK9-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store i64* [[TMP41]], i64** [[TMP51]], align 8
// CHECK9-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP52]], align 8
// CHECK9-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store i8** null, i8*** [[TMP53]], align 8
// CHECK9-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store i8** null, i8*** [[TMP54]], align 8
// CHECK9-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 [[ADD]], i64* [[TMP55]], align 8
// CHECK9-NEXT:    [[TMP56:%.*]] = 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_l81.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP57:%.*]] = icmp ne i32 [[TMP56]], 0
// CHECK9-NEXT:    br i1 [[TMP57]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK9:       omp_offload.failed:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81(i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP1]], i64 [[TMP3]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    [[TMP58:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef signext [[TMP58]])
// CHECK9-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK9-NEXT:    [[TMP59:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK9-NEXT:    call void @llvm.stackrestore(i8* [[TMP59]])
// CHECK9-NEXT:    [[TMP60:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK9-NEXT:    ret i32 [[TMP60]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
// CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP3]], i32* [[CONV4]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV3]], align 4
// CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP5]], i32* [[CONV5]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I13:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[J14:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV3]], align 4
// CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK9-NEXT:    [[CONV7:%.*]] = sext i32 [[DIV]] to i64
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP6]], 0
// CHECK9-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK9-NEXT:    [[CONV10:%.*]] = sext i32 [[DIV9]] to i64
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV7]], [[CONV10]]
// CHECK9-NEXT:    [[SUB11:%.*]] = sub nsw i64 [[MUL]], 1
// CHECK9-NEXT:    store i64 [[SUB11]], i64* [[DOTCAPTURE_EXPR_6]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[J]], align 4
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
// CHECK9-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK9:       land.lhs.true:
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[CMP12:%.*]] = icmp slt i32 0, [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP12]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
// CHECK9:       omp.precond.then:
// CHECK9-NEXT:    store i64 0, i64* [[DOTOMP_COMB_LB]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
// CHECK9-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_COMB_UB]], align 8
// CHECK9-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_COMB_LB]], i64* [[DOTOMP_COMB_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
// CHECK9-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
// CHECK9-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
// CHECK9-NEXT:    [[CMP15:%.*]] = icmp sgt i64 [[TMP12]], [[TMP13]]
// CHECK9-NEXT:    br i1 [[CMP15]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i64 [[COND]], i64* [[DOTOMP_COMB_UB]], align 8
// CHECK9-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_COMB_LB]], align 8
// CHECK9-NEXT:    store i64 [[TMP16]], i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
// CHECK9-NEXT:    [[CMP16:%.*]] = icmp sle i64 [[TMP17]], [[TMP18]]
// CHECK9-NEXT:    br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_COMB_LB]], align 8
// CHECK9-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[CONV17:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP21]], i32* [[CONV17]], align 4
// CHECK9-NEXT:    [[TMP22:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[CONV3]], align 4
// CHECK9-NEXT:    [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP23]], i32* [[CONV18]], align 4
// CHECK9-NEXT:    [[TMP24:%.*]] = load i64, i64* [[M_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 7, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, i64, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP19]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP24]], i64 [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP25]], [[TMP26]]
// CHECK9-NEXT:    store i64 [[ADD]], i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP28]])
// CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK9:       omp.precond.end:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I13:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[J14:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[M]], i64* [[M_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[M_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV3]], align 4
// CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK9-NEXT:    [[CONV7:%.*]] = sext i32 [[DIV]] to i64
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP6]], 0
// CHECK9-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
// CHECK9-NEXT:    [[CONV10:%.*]] = sext i32 [[DIV9]] to i64
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV7]], [[CONV10]]
// CHECK9-NEXT:    [[SUB11:%.*]] = sub nsw i64 [[MUL]], 1
// CHECK9-NEXT:    store i64 [[SUB11]], i64* [[DOTCAPTURE_EXPR_6]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[J]], align 4
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
// CHECK9-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK9:       land.lhs.true:
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[CMP12:%.*]] = icmp slt i32 0, [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP12]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
// CHECK9:       omp.precond.then:
// CHECK9-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
// CHECK9-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_UB]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK9-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[TMP10]], i64* [[DOTOMP_LB]], align 8
// CHECK9-NEXT:    store i64 [[TMP11]], i64* [[DOTOMP_UB]], align 8
// CHECK9-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
// CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK9-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
// CHECK9-NEXT:    [[CMP15:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
// CHECK9-NEXT:    br i1 [[CMP15]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
// CHECK9-NEXT:    store i64 [[TMP18]], i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK9-NEXT:    [[CMP16:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
// CHECK9-NEXT:    br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[SUB17:%.*]] = sub nsw i32 [[TMP22]], 0
// CHECK9-NEXT:    [[DIV18:%.*]] = sdiv i32 [[SUB17]], 1
// CHECK9-NEXT:    [[MUL19:%.*]] = mul nsw i32 1, [[DIV18]]
// CHECK9-NEXT:    [[CONV20:%.*]] = sext i32 [[MUL19]] to i64
// CHECK9-NEXT:    [[DIV21:%.*]] = sdiv i64 [[TMP21]], [[CONV20]]
// CHECK9-NEXT:    [[MUL22:%.*]] = mul nsw i64 [[DIV21]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL22]]
// CHECK9-NEXT:    [[CONV23:%.*]] = trunc i64 [[ADD]] to i32
// CHECK9-NEXT:    store i32 [[CONV23]], i32* [[I13]], align 4
// CHECK9-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[SUB24:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK9-NEXT:    [[DIV25:%.*]] = sdiv i32 [[SUB24]], 1
// CHECK9-NEXT:    [[MUL26:%.*]] = mul nsw i32 1, [[DIV25]]
// CHECK9-NEXT:    [[CONV27:%.*]] = sext i32 [[MUL26]] to i64
// CHECK9-NEXT:    [[DIV28:%.*]] = sdiv i64 [[TMP24]], [[CONV27]]
// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK9-NEXT:    [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
// CHECK9-NEXT:    [[MUL31:%.*]] = mul nsw i32 1, [[DIV30]]
// CHECK9-NEXT:    [[CONV32:%.*]] = sext i32 [[MUL31]] to i64
// CHECK9-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[DIV28]], [[CONV32]]
// CHECK9-NEXT:    [[SUB34:%.*]] = sub nsw i64 [[TMP23]], [[MUL33]]
// CHECK9-NEXT:    [[MUL35:%.*]] = mul nsw i64 [[SUB34]], 1
// CHECK9-NEXT:    [[ADD36:%.*]] = add nsw i64 0, [[MUL35]]
// CHECK9-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
// CHECK9-NEXT:    store i32 [[CONV37]], i32* [[J14]], align 4
// CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I13]], align 4
// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP27]] to i64
// CHECK9-NEXT:    [[TMP28:%.*]] = mul nsw i64 [[IDXPROM]], [[TMP1]]
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP28]]
// CHECK9-NEXT:    [[TMP29:%.*]] = load i32, i32* [[J14]], align 4
// CHECK9-NEXT:    [[IDXPROM38:%.*]] = sext i32 [[TMP29]] to i64
// CHECK9-NEXT:    [[ARRAYIDX39:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 [[IDXPROM38]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX39]], align 4
// 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:    [[TMP30:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    [[ADD40:%.*]] = add nsw i64 [[TMP30]], 1
// CHECK9-NEXT:    store i64 [[ADD40]], i64* [[DOTOMP_IV]], align 8
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP32]])
// CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK9:       omp.precond.end:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
// CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[A:%.*]] = alloca [10 x [2 x i32]], align 4
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x [2 x i32]]**
// CHECK9-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[TMP1]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x [2 x i32]]**
// CHECK9-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[TMP3]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK9-NEXT:    store i8* null, i8** [[TMP4]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 8
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP11]], align 8
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP12]], align 8
// CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store i8** null, i8*** [[TMP13]], align 8
// CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store i8** null, i8*** [[TMP14]], align 8
// CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 20, i64* [[TMP15]], align 8
// CHECK9-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]+}}__Z5tmainIiLi10ELi2EEiT__l68.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK9-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK9:       omp_offload.failed:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68([10 x [2 x i32]]* [[A]]) #[[ATTR3]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    ret i32 0
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68
// CHECK9-SAME: ([10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 8
// CHECK9-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 8
// CHECK9-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 [2 x i32]]*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), [10 x [2 x i32]]* [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 19, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK9-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)
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK9-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 [2 x i32]]*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x [2 x i32]]* [[TMP0]])
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK9-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 19, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK9-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK9-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)
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 19
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 2
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP12]], 2
// CHECK9-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 2
// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL5]]
// CHECK9-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
// CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 0, [[MUL6]]
// CHECK9-NEXT:    store i32 [[ADD7]], i32* [[J]], align 4
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], [10 x [2 x i32]]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[J]], align 4
// CHECK9-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP14]] to i64
// CHECK9-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[ARRAYIDX]], i64 0, i64 [[IDXPROM8]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX9]], align 4
// 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:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK9-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK9-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 4
// CHECK11-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[M:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK11-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK11-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
// CHECK11-NEXT:    store i32 100, i32* [[N]], align 4
// CHECK11-NEXT:    store i32 2, i32* [[M]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[M]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK11-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
// CHECK11-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP3]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR1]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[N_CASTED]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[M]], align 4
// CHECK11-NEXT:    store i32 [[TMP6]], i32* [[M_CASTED]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
// CHECK11-NEXT:    [[TMP9:%.*]] = mul nuw i32 [[TMP8]], 4
// CHECK11-NEXT:    [[TMP10:%.*]] = sext i32 [[TMP9]] to i64
// CHECK11-NEXT:    [[TMP11:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK11-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP11]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes to i8*), i32 40, i1 false)
// CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK11-NEXT:    store i32 [[TMP5]], i32* [[TMP13]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK11-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
// CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK11-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK11-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
// CHECK11-NEXT:    store i32 [[TMP7]], i32* [[TMP18]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK11-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK11-NEXT:    store i32 [[TMP7]], i32* [[TMP20]], align 4
// CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK11-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK11-NEXT:    store i32 [[TMP0]], i32* [[TMP23]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
// CHECK11-NEXT:    store i32 [[TMP0]], i32* [[TMP25]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK11-NEXT:    store i8* null, i8** [[TMP26]], align 4
// CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK11-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[TMP28]], align 4
// CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK11-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[TMP30]], align 4
// CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK11-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK11-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK11-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32**
// CHECK11-NEXT:    store i32* [[VLA]], i32** [[TMP33]], align 4
// CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK11-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32**
// CHECK11-NEXT:    store i32* [[VLA]], i32** [[TMP35]], align 4
// CHECK11-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK11-NEXT:    store i64 [[TMP10]], i64* [[TMP36]], align 4
// CHECK11-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK11-NEXT:    store i8* null, i8** [[TMP37]], align 4
// CHECK11-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP41:%.*]] = load i32, i32* [[N]], align 4
// CHECK11-NEXT:    store i32 [[TMP41]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP42:%.*]] = load i32, i32* [[M]], align 4
// CHECK11-NEXT:    store i32 [[TMP42]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK11-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP43]], 0
// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
// CHECK11-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK11-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[TMP44]], 0
// CHECK11-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
// CHECK11-NEXT:    [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
// CHECK11-NEXT:    [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
// CHECK11-NEXT:    store i64 [[SUB7]], i64* [[DOTCAPTURE_EXPR_3]], align 8
// CHECK11-NEXT:    [[TMP45:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP45]], 1
// CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 1, i32* [[TMP46]], align 4
// CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK11-NEXT:    store i32 5, i32* [[TMP47]], align 4
// CHECK11-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK11-NEXT:    store i8** [[TMP38]], i8*** [[TMP48]], align 4
// CHECK11-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK11-NEXT:    store i8** [[TMP39]], i8*** [[TMP49]], align 4
// CHECK11-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK11-NEXT:    store i64* [[TMP40]], i64** [[TMP50]], align 4
// CHECK11-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK11-NEXT:    store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP51]], align 4
// CHECK11-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK11-NEXT:    store i8** null, i8*** [[TMP52]], align 4
// CHECK11-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK11-NEXT:    store i8** null, i8*** [[TMP53]], align 4
// CHECK11-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK11-NEXT:    store i64 [[ADD]], i64* [[TMP54]], align 8
// CHECK11-NEXT:    [[TMP55:%.*]] = 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_l81.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK11-NEXT:    [[TMP56:%.*]] = icmp ne i32 [[TMP55]], 0
// CHECK11-NEXT:    br i1 [[TMP56]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK11:       omp_offload.failed:
// CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81(i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP0]], i32 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK11:       omp_offload.cont:
// CHECK11-NEXT:    [[TMP57:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
// CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef [[TMP57]])
// CHECK11-NEXT:    store i32 [[CALL]], i32* [[RETVAL]], align 4
// CHECK11-NEXT:    [[TMP58:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK11-NEXT:    call void @llvm.stackrestore(i8* [[TMP58]])
// CHECK11-NEXT:    [[TMP59:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP59]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
// CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], i32* [[M_CASTED]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I11:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[J12:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP6]], 0
// CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK11-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
// CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
// CHECK11-NEXT:    store i64 [[SUB9]], i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK11-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[J]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK11:       land.lhs.true:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP8]]
// CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
// CHECK11:       omp.precond.then:
// CHECK11-NEXT:    store i64 0, i64* [[DOTOMP_COMB_LB]], align 8
// CHECK11-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK11-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_COMB_UB]], align 8
// CHECK11-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_COMB_LB]], i64* [[DOTOMP_COMB_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
// CHECK11-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
// CHECK11-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK11-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP12]], [[TMP13]]
// CHECK11-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i64 [[COND]], i64* [[DOTOMP_COMB_UB]], align 8
// CHECK11-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_COMB_LB]], align 8
// CHECK11-NEXT:    store i64 [[TMP16]], i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
// CHECK11-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP17]], [[TMP18]]
// CHECK11-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_COMB_LB]], align 8
// CHECK11-NEXT:    [[TMP20:%.*]] = trunc i64 [[TMP19]] to i32
// CHECK11-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
// CHECK11-NEXT:    [[TMP22:%.*]] = trunc i64 [[TMP21]] to i32
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP23]], i32* [[N_CASTED]], align 4
// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP25]], i32* [[M_CASTED]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = load i32, i32* [[M_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 7, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, i32, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP20]], i32 [[TMP22]], i32 [[TMP24]], i32 [[TMP26]], i32 [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP27]], [[TMP28]]
// CHECK11-NEXT:    store i64 [[ADD]], i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK11:       omp.precond.end:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I13:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[J14:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[M]], i32* [[M_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[M_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP6]], 0
// CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
// CHECK11-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
// CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
// CHECK11-NEXT:    store i64 [[SUB9]], i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK11-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[J]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK11:       land.lhs.true:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP8]]
// CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
// CHECK11:       omp.precond.then:
// CHECK11-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
// CHECK11-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK11-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_UB]], align 8
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK11-NEXT:    [[CONV11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK11-NEXT:    [[CONV12:%.*]] = zext i32 [[TMP11]] to i64
// CHECK11-NEXT:    store i64 [[CONV11]], i64* [[DOTOMP_LB]], align 8
// CHECK11-NEXT:    store i64 [[CONV12]], i64* [[DOTOMP_UB]], align 8
// CHECK11-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
// CHECK11-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK11-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK11-NEXT:    [[CMP15:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
// CHECK11-NEXT:    br i1 [[CMP15]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
// CHECK11-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
// CHECK11-NEXT:    store i64 [[TMP18]], i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK11-NEXT:    [[CMP16:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
// CHECK11-NEXT:    br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[SUB17:%.*]] = sub nsw i32 [[TMP22]], 0
// CHECK11-NEXT:    [[DIV18:%.*]] = sdiv i32 [[SUB17]], 1
// CHECK11-NEXT:    [[MUL19:%.*]] = mul nsw i32 1, [[DIV18]]
// CHECK11-NEXT:    [[CONV20:%.*]] = sext i32 [[MUL19]] to i64
// CHECK11-NEXT:    [[DIV21:%.*]] = sdiv i64 [[TMP21]], [[CONV20]]
// CHECK11-NEXT:    [[MUL22:%.*]] = mul nsw i64 [[DIV21]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL22]]
// CHECK11-NEXT:    [[CONV23:%.*]] = trunc i64 [[ADD]] to i32
// CHECK11-NEXT:    store i32 [[CONV23]], i32* [[I13]], align 4
// CHECK11-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[SUB24:%.*]] = sub nsw i32 [[TMP25]], 0
// CHECK11-NEXT:    [[DIV25:%.*]] = sdiv i32 [[SUB24]], 1
// CHECK11-NEXT:    [[MUL26:%.*]] = mul nsw i32 1, [[DIV25]]
// CHECK11-NEXT:    [[CONV27:%.*]] = sext i32 [[MUL26]] to i64
// CHECK11-NEXT:    [[DIV28:%.*]] = sdiv i64 [[TMP24]], [[CONV27]]
// CHECK11-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK11-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK11-NEXT:    [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
// CHECK11-NEXT:    [[MUL31:%.*]] = mul nsw i32 1, [[DIV30]]
// CHECK11-NEXT:    [[CONV32:%.*]] = sext i32 [[MUL31]] to i64
// CHECK11-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[DIV28]], [[CONV32]]
// CHECK11-NEXT:    [[SUB34:%.*]] = sub nsw i64 [[TMP23]], [[MUL33]]
// CHECK11-NEXT:    [[MUL35:%.*]] = mul nsw i64 [[SUB34]], 1
// CHECK11-NEXT:    [[ADD36:%.*]] = add nsw i64 0, [[MUL35]]
// CHECK11-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
// CHECK11-NEXT:    store i32 [[CONV37]], i32* [[J14]], align 4
// CHECK11-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I13]], align 4
// CHECK11-NEXT:    [[TMP28:%.*]] = mul nsw i32 [[TMP27]], [[TMP1]]
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP28]]
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32, i32* [[J14]], align 4
// CHECK11-NEXT:    [[ARRAYIDX38:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i32 [[TMP29]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX38]], align 4
// 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:    [[TMP30:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    [[ADD39:%.*]] = add nsw i64 [[TMP30]], 1
// CHECK11-NEXT:    store i64 [[ADD39]], i64* [[DOTOMP_IV]], align 8
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP32]])
// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK11:       omp.precond.end:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
// CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A:%.*]] = alloca [10 x [2 x i32]], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x [2 x i32]]**
// CHECK11-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[TMP1]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x [2 x i32]]**
// CHECK11-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[TMP3]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK11-NEXT:    store i8* null, i8** [[TMP4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK11-NEXT:    store i32 1, i32* [[TMP7]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK11-NEXT:    store i32 1, i32* [[TMP8]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK11-NEXT:    store i8** [[TMP5]], i8*** [[TMP9]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK11-NEXT:    store i8** [[TMP6]], i8*** [[TMP10]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK11-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP11]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK11-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP12]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK11-NEXT:    store i8** null, i8*** [[TMP13]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK11-NEXT:    store i8** null, i8*** [[TMP14]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK11-NEXT:    store i64 20, i64* [[TMP15]], align 8
// CHECK11-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]+}}__Z5tmainIiLi10ELi2EEiT__l68.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK11-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK11-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK11:       omp_offload.failed:
// CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68([10 x [2 x i32]]* [[A]]) #[[ATTR3]]
// CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK11:       omp_offload.cont:
// CHECK11-NEXT:    ret i32 0
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68
// CHECK11-SAME: ([10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 4
// CHECK11-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 4
// CHECK11-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 [2 x i32]]*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), [10 x [2 x i32]]* [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    store i32 19, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK11-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)
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-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 [2 x i32]]*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x [2 x i32]]* [[TMP0]])
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK11-NEXT:    store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 19, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK11-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)
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 19
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 2
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP12]], 2
// CHECK11-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL4]]
// CHECK11-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
// CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
// CHECK11-NEXT:    store i32 [[ADD6]], i32* [[J]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], [10 x [2 x i32]]* [[TMP0]], i32 0, i32 [[TMP13]]
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[J]], align 4
// CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[ARRAYIDX]], i32 0, i32 [[TMP14]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX7]], align 4
// 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:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK11-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK11-NEXT:    ret void
//