xref: /llvm-project-15.0.7/clang/test/OpenMP/reduction_implicit_map.cpp (revision 5300263c)

// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-cuda-mode -x c++ \
// RUN:  -triple powerpc64le-unknown-unknown -DCUDA \
// RUN:  -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm-bc %s -o \
// RUN:  %t-ppc-host.bc

// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-cuda-mode -x c++ \
// RUN:  -triple nvptx64-unknown-unknown -DCUA \
// RUN:  -fopenmp-targets=nvptx64-nvidia-cuda -DCUDA -emit-llvm %s \
// RUN:  -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc \
// RUN:  -o - | FileCheck %s --check-prefix CHECK

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

// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ \
// RUN:   -triple i386-unknown-unknown \
// RUN:   -fopenmp-targets=i386-pc-linux-gnu -emit-llvm \
// RUN:   %s -o - | FileCheck  %s \
// RUN:   --check-prefix=CHECK2


#if defined(CUDA)
// expected-no-diagnostics

int foo(int n) {
  double *e;
  //no error and no implicit map generated for e[:1]
  #pragma omp target parallel reduction(+: e[:1])
    *e=10;
  ;
  return 0;
}
// CHECK-NOT @.offload_maptypes
#elif defined(DIAG)
class S2 {
  mutable int a;
public:
  S2():a(0) { }
  S2(S2 &s2):a(s2.a) { }
  S2 &operator +(S2 &s);
};
int bar() {
 S2 o[5];
  //warnig "copyable and not guaranteed to be mapped correctly" and
  //implicit map generated.
#pragma omp target parallel reduction(+:o[0]) //expected-warning {{Type 'S2' is not trivially copyable and not guaranteed to be mapped correctly}}
  for (int i = 0; i < 10; i++);
  double b[10][10][10];
  //no error no implicit map generated, the map for b is generated but not
  //for b[0:2][2:4][1].
#pragma omp target parallel for reduction(task, +: b[0:2][2:4][1])
  for (long long i = 0; i < 10; ++i);
  return 0;
}
// map for variable o
// map for b:
#else
// expected-no-diagnostics

// generate implicit map for array elements or array sections in reduction
// clause. In following case: the implicit map is generate for output[0]
// with map size 4 and output[:3] with map size 12.
void sum(int* input, int size, int* output)
{
#pragma omp target teams distribute parallel for reduction(+: output[0]) \
                                                 map(to: input [0:size])
  for (int i = 0; i < size; i++)
    output[0] += input[i];
#pragma omp target teams distribute parallel for reduction(+: output[:3])  \
                                                 map(to: input [0:size])
  for (int i = 0; i < size; i++)
    output[0] += input[i];
  int a[10];
#pragma omp target parallel reduction(+: a[:2])
  for (int i = 0; i < size; i++)
    ;
#pragma omp target parallel reduction(+: a[3])
  for (int i = 0; i < size; i++)
    ;
}
#endif
int main()
{
#if defined(CUDA)
  int a = foo(10);
#elif defined(DIAG)
  int a = bar();
#else
  const int size = 100;
  int *array = new int[size];
  int result = 0;
  sum(array, size, &result);
#endif
  return 0;
}
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l32
// CHECK-SAME: (double* noundef [[E:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 8
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x i8*], align 8
// CHECK-NEXT:    store double* [[E]], double** [[E_ADDR]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* @[[GLOB1:[0-9]+]], i8 2, i1 false, i1 true)
// CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK:       user_code.entry:
// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
// CHECK-NEXT:    [[TMP2:%.*]] = load double*, double** [[E_ADDR]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-NEXT:    [[TMP4:%.*]] = bitcast double* [[TMP2]] to i8*
// CHECK-NEXT:    store i8* [[TMP4]], i8** [[TMP3]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = bitcast [1 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
// CHECK-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, double*)* @__omp_outlined__ to i8*), i8* null, i8** [[TMP5]], i64 1)
// CHECK-NEXT:    call void @__kmpc_target_deinit(%struct.ident_t* @[[GLOB1]], i8 2, i1 true)
// CHECK-NEXT:    ret void
// CHECK:       worker.exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@__omp_outlined__
// CHECK-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], double* noundef [[E:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK-NEXT:    [[E_ADDR:%.*]] = alloca double*, align 8
// CHECK-NEXT:    [[E2:%.*]] = alloca double, align 8
// CHECK-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK-NEXT:    store double* [[E]], double** [[E_ADDR]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load double*, double** [[E_ADDR]], align 8
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP0]], i64 0
// CHECK-NEXT:    [[TMP1:%.*]] = load double*, double** [[E_ADDR]], align 8
// CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds double, double* [[TMP1]], i64 0
// CHECK-NEXT:    store double 0.000000e+00, double* [[E2]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load double*, double** [[E_ADDR]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = ptrtoint double* [[TMP2]] to i64
// CHECK-NEXT:    [[TMP4:%.*]] = ptrtoint double* [[ARRAYIDX]] to i64
// CHECK-NEXT:    [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
// CHECK-NEXT:    [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (double* getelementptr (double, double* null, i32 1) to i64)
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr double, double* [[E2]], i64 [[TMP6]]
// CHECK-NEXT:    store double* [[TMP7]], double** [[TMP]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load double*, double** [[TMP]], align 8
// CHECK-NEXT:    store double 1.000000e+01, double* [[TMP8]], align 8
// CHECK-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP12:%.*]] = bitcast double* [[E2]] to i8*
// CHECK-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 8
// CHECK-NEXT:    [[TMP13:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1, i64 8, i8* [[TMP13]], void (i8*, i16, i16, i16)* @_omp_reduction_shuffle_and_reduce_func, void (i8*, i32)* @_omp_reduction_inter_warp_copy_func)
// CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 1
// CHECK-NEXT:    br i1 [[TMP15]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP16:%.*]] = load double, double* [[ARRAYIDX]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load double, double* [[E2]], align 8
// CHECK-NEXT:    [[ADD:%.*]] = fadd double [[TMP16]], [[TMP17]]
// CHECK-NEXT:    store double [[ADD]], double* [[ARRAYIDX]], align 8
// CHECK-NEXT:    call void @__kmpc_nvptx_end_reduce_nowait(i32 [[TMP10]])
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func
// CHECK-SAME: (i8* noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8
// CHECK-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], i16* [[DOTADDR1]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], i16* [[DOTADDR2]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], i16* [[DOTADDR3]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, i16* [[DOTADDR1]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTADDR2]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = load i16, i16* [[DOTADDR3]], align 2
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to double**
// CHECK-NEXT:    [[TMP11:%.*]] = load double*, double** [[TMP10]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = getelementptr double, double* [[TMP11]], i64 1
// CHECK-NEXT:    [[TMP14:%.*]] = bitcast double* [[TMP13]] to i8*
// CHECK-NEXT:    [[TMP15:%.*]] = bitcast double* [[TMP11]] to i64*
// CHECK-NEXT:    [[TMP16:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i64*
// CHECK-NEXT:    [[TMP17:%.*]] = load i64, i64* [[TMP15]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP19:%.*]] = trunc i32 [[TMP18]] to i16
// CHECK-NEXT:    [[TMP20:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP17]], i16 [[TMP7]], i16 [[TMP19]])
// CHECK-NEXT:    store i64 [[TMP20]], i64* [[TMP16]], align 8
// CHECK-NEXT:    [[TMP21:%.*]] = getelementptr i64, i64* [[TMP15]], i64 1
// CHECK-NEXT:    [[TMP22:%.*]] = getelementptr i64, i64* [[TMP16]], i64 1
// CHECK-NEXT:    [[TMP23:%.*]] = bitcast double* [[DOTOMP_REDUCTION_ELEMENT]] to i8*
// CHECK-NEXT:    store i8* [[TMP23]], i8** [[TMP12]], align 8
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP8]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i16 [[TMP8]], 1
// CHECK-NEXT:    [[TMP26:%.*]] = icmp ult i16 [[TMP6]], [[TMP7]]
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = icmp eq i16 [[TMP8]], 2
// CHECK-NEXT:    [[TMP29:%.*]] = and i16 [[TMP6]], 1
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP29]], 0
// CHECK-NEXT:    [[TMP31:%.*]] = and i1 [[TMP28]], [[TMP30]]
// CHECK-NEXT:    [[TMP32:%.*]] = icmp sgt i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP33:%.*]] = and i1 [[TMP31]], [[TMP32]]
// CHECK-NEXT:    [[TMP34:%.*]] = or i1 [[TMP24]], [[TMP27]]
// CHECK-NEXT:    [[TMP35:%.*]] = or i1 [[TMP34]], [[TMP33]]
// CHECK-NEXT:    br i1 [[TMP35]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[TMP5]] to i8*
// CHECK-NEXT:    [[TMP37:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to i8*
// CHECK-NEXT:    call void @"_omp$reduction$reduction_func"(i8* [[TMP36]], i8* [[TMP37]]) #[[ATTR3:[0-9]+]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP38:%.*]] = icmp eq i16 [[TMP8]], 1
// CHECK-NEXT:    [[TMP39:%.*]] = icmp uge i16 [[TMP6]], [[TMP7]]
// CHECK-NEXT:    [[TMP40:%.*]] = and i1 [[TMP38]], [[TMP39]]
// CHECK-NEXT:    br i1 [[TMP40]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to double**
// CHECK-NEXT:    [[TMP43:%.*]] = load double*, double** [[TMP42]], align 8
// CHECK-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to double**
// CHECK-NEXT:    [[TMP46:%.*]] = load double*, double** [[TMP45]], align 8
// CHECK-NEXT:    [[TMP47:%.*]] = load double, double* [[TMP43]], align 8
// CHECK-NEXT:    store double [[TMP47]], double* [[TMP46]], align 8
// CHECK-NEXT:    br label [[IFCONT6:%.*]]
// CHECK:       else5:
// CHECK-NEXT:    br label [[IFCONT6]]
// CHECK:       ifcont6:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func
// CHECK-SAME: (i8* noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR2]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
// CHECK-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
// CHECK-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], i32* [[DOTADDR1]], align 4
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP4]], 31
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP5]], 5
// CHECK-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP6]] to [1 x i8*]*
// CHECK-NEXT:    store i32 0, i32* [[DOTCNT_ADDR]], align 4
// CHECK-NEXT:    br label [[PRECOND:%.*]]
// CHECK:       precond:
// CHECK-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCNT_ADDR]], align 4
// CHECK-NEXT:    [[TMP9:%.*]] = icmp ult i32 [[TMP8]], 2
// CHECK-NEXT:    br i1 [[TMP9]], label [[BODY:%.*]], label [[EXIT:%.*]]
// CHECK:       body:
// CHECK-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP2]])
// CHECK-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
// CHECK-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP7]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[TMP10]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = bitcast i8* [[TMP11]] to i32*
// CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[TMP12]], i32 [[TMP8]]
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP15]], i32 addrspace(3)* [[TMP14]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
// CHECK-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTADDR1]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP3]], [[TMP16]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN2:%.*]], label [[ELSE3:%.*]]
// CHECK:       then2:
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace(3)* @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP3]]
// CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP7]], i64 0, i64 0
// CHECK-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to i32*
// CHECK-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[TMP20]], i32 [[TMP8]]
// CHECK-NEXT:    [[TMP22:%.*]] = load volatile i32, i32 addrspace(3)* [[TMP17]], align 4
// CHECK-NEXT:    store i32 [[TMP22]], i32* [[TMP21]], align 4
// CHECK-NEXT:    br label [[IFCONT4:%.*]]
// CHECK:       else3:
// CHECK-NEXT:    br label [[IFCONT4]]
// CHECK:       ifcont4:
// CHECK-NEXT:    [[TMP23:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-NEXT:    store i32 [[TMP23]], i32* [[DOTCNT_ADDR]], align 4
// CHECK-NEXT:    br label [[PRECOND]]
// CHECK:       exit:
// CHECK-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z3barv
// CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[O:%.*]] = alloca [5 x %class.S2], align 4
// 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:    [[B:%.*]] = alloca [10 x [10 x [10 x double]]], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [5 x %class.S2], [5 x %class.S2]* [[O]], i32 0, i32 0
// CHECK1-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[CLASS_S2:%.*]], %class.S2* [[ARRAY_BEGIN]], i64 5
// CHECK1-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK1:       arrayctor.loop:
// CHECK1-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi %class.S2* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK1-NEXT:    call void @_ZN2S2C1Ev(%class.S2* noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK1-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[CLASS_S2]], %class.S2* [[ARRAYCTOR_CUR]], i64 1
// CHECK1-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq %class.S2* [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK1-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK1:       arrayctor.cont:
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [5 x %class.S2], [5 x %class.S2]* [[O]], i64 0, i64 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 [5 x %class.S2]**
// CHECK1-NEXT:    store [5 x %class.S2]* [[O]], [5 x %class.S2]** [[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 %class.S2**
// CHECK1-NEXT:    store %class.S2* [[ARRAYIDX]], %class.S2** [[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 0, i64* [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 1, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barv_l50.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]+}}__Z3barv_l50([5 x %class.S2]* [[O]]) #[[ATTR8:[0-9]+]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x [10 x [10 x double]]]**
// CHECK1-NEXT:    store [10 x [10 x [10 x double]]]* [[B]], [10 x [10 x [10 x double]]]** [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x [10 x [10 x double]]]**
// CHECK1-NEXT:    store [10 x [10 x [10 x double]]]* [[B]], [10 x [10 x [10 x double]]]** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS4:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP25]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 1
// CHECK1-NEXT:    store i32 1, i32* [[TMP26]], align 4
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP23]], i8*** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP24]], i8*** [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.3, i32 0, i32 0), i64** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.4, i32 0, i32 0), i64** [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 8
// CHECK1-NEXT:    store i64 0, i64* [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 1, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barv_l55.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]])
// CHECK1-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
// CHECK1-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
// CHECK1:       omp_offload.failed5:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barv_l55([10 x [10 x [10 x double]]]* [[B]]) #[[ATTR8]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
// CHECK1:       omp_offload.cont6:
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN2S2C1Ev
// CHECK1-SAME: (%class.S2* noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %class.S2*, align 8
// CHECK1-NEXT:    store %class.S2* [[THIS]], %class.S2** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %class.S2*, %class.S2** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN2S2C2Ev(%class.S2* noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barv_l50
// CHECK1-SAME: ([5 x %class.S2]* noundef nonnull align 4 dereferenceable(20) [[O:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[O_ADDR:%.*]] = alloca [5 x %class.S2]*, align 8
// CHECK1-NEXT:    store [5 x %class.S2]* [[O]], [5 x %class.S2]** [[O_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [5 x %class.S2]*, [5 x %class.S2]** [[O_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [5 x %class.S2]*)* @.omp_outlined. to void (i32*, i32*, ...)*), [5 x %class.S2]* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [5 x %class.S2]* noundef nonnull align 4 dereferenceable(20) [[O:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[O_ADDR:%.*]] = alloca [5 x %class.S2]*, align 8
// CHECK1-NEXT:    [[O1:%.*]] = alloca [[CLASS_S2:%.*]], align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store [5 x %class.S2]* [[O]], [5 x %class.S2]** [[O_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [5 x %class.S2]*, [5 x %class.S2]** [[O_ADDR]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [5 x %class.S2], [5 x %class.S2]* [[TMP0]], i64 0, i64 0
// CHECK1-NEXT:    call void @_ZN2S2C1Ev(%class.S2* noundef nonnull align 4 dereferenceable(4) [[O1]])
// CHECK1-NEXT:    [[TMP1:%.*]] = bitcast [5 x %class.S2]* [[TMP0]] to %class.S2*
// CHECK1-NEXT:    [[TMP2:%.*]] = ptrtoint %class.S2* [[TMP1]] to i64
// CHECK1-NEXT:    [[TMP3:%.*]] = ptrtoint %class.S2* [[ARRAYIDX]] to i64
// CHECK1-NEXT:    [[TMP4:%.*]] = sub i64 [[TMP2]], [[TMP3]]
// CHECK1-NEXT:    [[TMP5:%.*]] = sdiv exact i64 [[TMP4]], ptrtoint (%class.S2* getelementptr ([[CLASS_S2]], %class.S2* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr [[CLASS_S2]], %class.S2* [[O1]], i64 [[TMP5]]
// CHECK1-NEXT:    [[TMP7:%.*]] = bitcast %class.S2* [[TMP6]] to [5 x %class.S2]*
// CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
// CHECK1:       for.cond:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP8]], 10
// CHECK1-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK1:       for.body:
// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
// CHECK1:       for.inc:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
// CHECK1:       for.end:
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %class.S2* [[O1]] to i8*
// CHECK1-NEXT:    store i8* [[TMP11]], i8** [[TMP10]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP13]], i32 1, i64 8, i8* [[TMP14]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) %class.S2* @_ZN2S2plERS_(%class.S2* noundef nonnull align 4 dereferenceable(4) [[ARRAYIDX]], %class.S2* noundef nonnull align 4 dereferenceable(4) [[O1]])
// CHECK1-NEXT:    [[TMP16:%.*]] = bitcast %class.S2* [[ARRAYIDX]] to i8*
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast %class.S2* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP16]], i8* align 4 [[TMP17]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP19]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL2:%.*]] = call noundef nonnull align 4 dereferenceable(4) %class.S2* @_ZN2S2plERS_(%class.S2* noundef nonnull align 4 dereferenceable(4) [[ARRAYIDX]], %class.S2* noundef nonnull align 4 dereferenceable(4) [[O1]])
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast %class.S2* [[ARRAYIDX]] to i8*
// CHECK1-NEXT:    [[TMP21:%.*]] = bitcast %class.S2* [[CALL2]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP20]], i8* align 4 [[TMP21]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP19]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
// CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %class.S2*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %class.S2*
// CHECK1-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) %class.S2* @_ZN2S2plERS_(%class.S2* noundef nonnull align 4 dereferenceable(4) [[TMP11]], %class.S2* noundef nonnull align 4 dereferenceable(4) [[TMP8]])
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast %class.S2* [[TMP11]] to i8*
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %class.S2* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i64 4, i1 false)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barv_l55
// CHECK1-SAME: ([10 x [10 x [10 x double]]]* noundef nonnull align 8 dereferenceable(8000) [[B:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x [10 x [10 x double]]]*, align 8
// CHECK1-NEXT:    store [10 x [10 x [10 x double]]]* [[B]], [10 x [10 x [10 x double]]]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x [10 x [10 x double]]]*, [10 x [10 x [10 x double]]]** [[B_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x [10 x [10 x double]]]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x [10 x [10 x double]]]* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x [10 x [10 x double]]]* noundef nonnull align 8 dereferenceable(8000) [[B:%.*]]) #[[ATTR3]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x [10 x [10 x double]]]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [1 x %struct.kmp_taskred_input_t], align 8
// CHECK1-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store [10 x [10 x [10 x double]]]* [[B]], [10 x [10 x [10 x double]]]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x [10 x [10 x double]]]*, [10 x [10 x [10 x double]]]** [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
// CHECK1-NEXT:    store i64 9, i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x [10 x double]]], [10 x [10 x [10 x double]]]* [[TMP0]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [10 x [10 x double]], [10 x [10 x double]]* [[ARRAYIDX]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYDECAY]], i64 2
// CHECK1-NEXT:    [[ARRAYDECAY2:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX1]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds double, double* [[ARRAYDECAY2]], i64 1
// CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [10 x [10 x [10 x double]]], [10 x [10 x [10 x double]]]* [[TMP0]], i64 0, i64 1
// CHECK1-NEXT:    [[ARRAYDECAY5:%.*]] = getelementptr inbounds [10 x [10 x double]], [10 x [10 x double]]* [[ARRAYIDX4]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYDECAY5]], i64 5
// CHECK1-NEXT:    [[ARRAYDECAY7:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX6]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds double, double* [[ARRAYDECAY7]], i64 1
// CHECK1-NEXT:    [[TMP1:%.*]] = ptrtoint double* [[ARRAYIDX8]] to i64
// CHECK1-NEXT:    [[TMP2:%.*]] = ptrtoint double* [[ARRAYIDX3]] to i64
// CHECK1-NEXT:    [[TMP3:%.*]] = sub i64 [[TMP1]], [[TMP2]]
// CHECK1-NEXT:    [[TMP4:%.*]] = sdiv exact i64 [[TMP3]], ptrtoint (double* getelementptr (double, double* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP5:%.*]] = add nuw i64 [[TMP4]], 1
// CHECK1-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], ptrtoint (double* getelementptr (double, double* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave()
// CHECK1-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    [[VLA:%.*]] = alloca double, i64 [[TMP5]], align 8
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR0]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr double, double* [[VLA]], i64 [[TMP5]]
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq double* [[VLA]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi double* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    store double 0.000000e+00, double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 8
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq double* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP9:%.*]] = bitcast [10 x [10 x [10 x double]]]* [[TMP0]] to double*
// CHECK1-NEXT:    [[TMP10:%.*]] = ptrtoint double* [[TMP9]] to i64
// CHECK1-NEXT:    [[TMP11:%.*]] = ptrtoint double* [[ARRAYIDX3]] to i64
// CHECK1-NEXT:    [[TMP12:%.*]] = sub i64 [[TMP10]], [[TMP11]]
// CHECK1-NEXT:    [[TMP13:%.*]] = sdiv exact i64 [[TMP12]], ptrtoint (double* getelementptr (double, double* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr double, double* [[VLA]], i64 [[TMP13]]
// CHECK1-NEXT:    [[TMP15:%.*]] = bitcast double* [[TMP14]] to [10 x [10 x [10 x double]]]*
// CHECK1-NEXT:    [[DOTRD_INPUT_GEP_:%.*]] = getelementptr inbounds [1 x %struct.kmp_taskred_input_t], [1 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
// CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [10 x [10 x [10 x double]]], [10 x [10 x [10 x double]]]* [[TMP0]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYDECAY10:%.*]] = getelementptr inbounds [10 x [10 x double]], [10 x [10 x double]]* [[ARRAYIDX9]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYDECAY10]], i64 2
// CHECK1-NEXT:    [[ARRAYDECAY12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[ARRAYDECAY12]], i64 1
// CHECK1-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [10 x [10 x [10 x double]]], [10 x [10 x [10 x double]]]* [[TMP0]], i64 0, i64 1
// CHECK1-NEXT:    [[ARRAYDECAY15:%.*]] = getelementptr inbounds [10 x [10 x double]], [10 x [10 x double]]* [[ARRAYIDX14]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYDECAY15]], i64 5
// CHECK1-NEXT:    [[ARRAYDECAY17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYDECAY17]], i64 1
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast double* [[VLA]] to i8*
// CHECK1-NEXT:    store i8* [[TMP17]], i8** [[TMP16]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP19:%.*]] = bitcast double* [[ARRAYIDX13]] to i8*
// CHECK1-NEXT:    store i8* [[TMP19]], i8** [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = ptrtoint double* [[ARRAYIDX18]] to i64
// CHECK1-NEXT:    [[TMP21:%.*]] = ptrtoint double* [[ARRAYIDX13]] to i64
// CHECK1-NEXT:    [[TMP22:%.*]] = sub i64 [[TMP20]], [[TMP21]]
// CHECK1-NEXT:    [[TMP23:%.*]] = sdiv exact i64 [[TMP22]], ptrtoint (double* getelementptr (double, double* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP24:%.*]] = add nuw i64 [[TMP23]], 1
// CHECK1-NEXT:    [[TMP25:%.*]] = mul nuw i64 [[TMP24]], ptrtoint (double* getelementptr (double, double* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
// CHECK1-NEXT:    store i64 [[TMP25]], i64* [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
// CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
// CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
// CHECK1-NEXT:    store i32 1, i32* [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast [1 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
// CHECK1-NEXT:    [[TMP34:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB2]], i32 [[TMP32]], i32 1, i32 1, i8* [[TMP33]])
// CHECK1-NEXT:    store i8* [[TMP34]], i8** [[DOTTASK_RED_]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP36]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
// CHECK1-NEXT:    [[TMP37:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP37]], 9
// 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:    [[TMP38:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP38]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    [[TMP39:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
// CHECK1-NEXT:    store i64 [[TMP39]], i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP40:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP41:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    [[CMP19:%.*]] = icmp sle i64 [[TMP40]], [[TMP41]]
// CHECK1-NEXT:    br i1 [[CMP19]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP42:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP42]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL]]
// CHECK1-NEXT:    store i64 [[ADD]], i64* [[I]], align 8
// 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:    [[TMP43:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP43]], 1
// CHECK1-NEXT:    store i64 [[ADD20]], i64* [[DOTOMP_IV]], align 8
// 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:    [[TMP44:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP45:%.*]] = load i32, i32* [[TMP44]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP45]])
// CHECK1-NEXT:    [[TMP46:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP47:%.*]] = load i32, i32* [[TMP46]], align 4
// CHECK1-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP47]], i32 1)
// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP49:%.*]] = bitcast double* [[VLA]] to i8*
// CHECK1-NEXT:    store i8* [[TMP49]], i8** [[TMP48]], align 8
// CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP51:%.*]] = inttoptr i64 [[TMP5]] to i8*
// CHECK1-NEXT:    store i8* [[TMP51]], i8** [[TMP50]], align 8
// CHECK1-NEXT:    [[TMP52:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP53:%.*]] = load i32, i32* [[TMP52]], align 4
// CHECK1-NEXT:    [[TMP54:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP55:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB1]], i32 [[TMP53]], i32 1, i64 16, i8* [[TMP54]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP55]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr double, double* [[ARRAYIDX3]], i64 [[TMP5]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq double* [[ARRAYIDX3]], [[TMP56]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE25:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi double* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST21:%.*]] = phi double* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT23:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP57:%.*]] = load double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], align 8
// CHECK1-NEXT:    [[TMP58:%.*]] = load double, double* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 8
// CHECK1-NEXT:    [[ADD22:%.*]] = fadd double [[TMP57]], [[TMP58]]
// CHECK1-NEXT:    store double [[ADD22]], double* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], align 8
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT23]] = getelementptr double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr double, double* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE24:%.*]] = icmp eq double* [[OMP_ARRAYCPY_DEST_ELEMENT23]], [[TMP56]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE24]], label [[OMP_ARRAYCPY_DONE25]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done25:
// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB1]], i32 [[TMP53]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr double, double* [[ARRAYIDX3]], i64 [[TMP5]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY26:%.*]] = icmp eq double* [[ARRAYIDX3]], [[TMP59]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY26]], label [[OMP_ARRAYCPY_DONE33:%.*]], label [[OMP_ARRAYCPY_BODY27:%.*]]
// CHECK1:       omp.arraycpy.body27:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST28:%.*]] = phi double* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT31:%.*]], [[OMP_ARRAYCPY_BODY27]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST29:%.*]] = phi double* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT30:%.*]], [[OMP_ARRAYCPY_BODY27]] ]
// CHECK1-NEXT:    [[TMP60:%.*]] = load double, double* [[OMP_ARRAYCPY_SRCELEMENTPAST28]], align 8
// CHECK1-NEXT:    [[TMP61:%.*]] = atomicrmw fadd double* [[OMP_ARRAYCPY_DESTELEMENTPAST29]], double [[TMP60]] monotonic, align 8
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT30]] = getelementptr double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST29]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT31]] = getelementptr double, double* [[OMP_ARRAYCPY_SRCELEMENTPAST28]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE32:%.*]] = icmp eq double* [[OMP_ARRAYCPY_DEST_ELEMENT30]], [[TMP59]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_DONE33]], label [[OMP_ARRAYCPY_BODY27]]
// CHECK1:       omp.arraycpy.done33:
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[TMP62:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP62]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.red_init.
// CHECK1-SAME: (i8* noalias noundef [[TMP0:%.*]], i8* noalias noundef [[TMP1:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = bitcast i8** [[DOTADDR]] to double**
// CHECK1-NEXT:    [[TMP3:%.*]] = load double*, double** [[TMP2]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr double, double* [[TMP3]], i64 [[TMP4]]
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq double* [[TMP3]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi double* [ [[TMP3]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    store double 0.000000e+00, double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 8
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq double* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.red_comb.
// CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8** [[DOTADDR]] to double**
// CHECK1-NEXT:    [[TMP4:%.*]] = load double*, double** [[TMP3]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8** [[DOTADDR1]] to double**
// CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[TMP5]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr double, double* [[TMP4]], i64 [[TMP2]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq double* [[TMP4]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi double* [ [[TMP6]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi double* [ [[TMP4]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP8:%.*]] = load double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = load double, double* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 8
// CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    store double [[ADD]], double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 8
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr double, double* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq double* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2
// CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to double*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to double*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr double, double* [[TMP11]], i64 [[TMP14]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq double* [[TMP11]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi double* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi double* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP16:%.*]] = load double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = load double, double* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 8
// CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[TMP16]], [[TMP17]]
// CHECK1-NEXT:    store double [[ADD]], double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 8
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr double, double* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr double, double* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq double* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN2S2C2Ev
// CHECK1-SAME: (%class.S2* noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %class.S2*, align 8
// CHECK1-NEXT:    store %class.S2* [[THIS]], %class.S2** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %class.S2*, %class.S2** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[CLASS_S2:%.*]], %class.S2* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@main
// CHECK1-SAME: () #[[ATTR10:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3barv()
// CHECK1-NEXT:    store i32 [[CALL]], i32* [[A]], align 4
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR11:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z3sumPiiS_
// CHECK2-SAME: (i32* noundef [[INPUT:%.*]], i32 noundef [[SIZE:%.*]], i32* noundef [[OUTPUT:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[INPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[SIZE_CASTED:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SIZE_CASTED4:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [3 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [3 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [3 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_SIZES10:%.*]] = alloca [3 x i64], align 4
// CHECK2-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
// CHECK2-NEXT:    [[SIZE_CASTED21:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS23:%.*]] = alloca [2 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS24:%.*]] = alloca [2 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS25:%.*]] = alloca [2 x i8*], align 4
// CHECK2-NEXT:    [[SIZE_CASTED29:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS31:%.*]] = alloca [2 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS32:%.*]] = alloca [2 x i8*], align 4
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS33:%.*]] = alloca [2 x i8*], align 4
// CHECK2-NEXT:    store i32* [[INPUT]], i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[OUTPUT]], i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 0
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, i32* [[TMP7]], i32 0
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = mul nuw i32 [[TMP8]], 4
// CHECK2-NEXT:    [[TMP10:%.*]] = sext i32 [[TMP9]] to i64
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP11]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[TMP13]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[TMP15]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
// CHECK2-NEXT:    store i32* [[TMP4]], i32** [[TMP18]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
// CHECK2-NEXT:    store i32* [[ARRAYIDX]], i32** [[TMP20]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32**
// CHECK2-NEXT:    store i32* [[TMP6]], i32** [[TMP23]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32**
// CHECK2-NEXT:    store i32* [[ARRAYIDX1]], i32** [[TMP25]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK2-NEXT:    store i64 [[TMP10]], i64* [[TMP26]], align 4
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP27]], align 4
// CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP33]], 1
// CHECK2-NEXT:    [[TMP34:%.*]] = zext i32 [[ADD]] to i64
// CHECK2-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP35]], align 4
// CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK2-NEXT:    store i32 3, i32* [[TMP36]], align 4
// CHECK2-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP28]], i8*** [[TMP37]], align 4
// CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP29]], i8*** [[TMP38]], align 4
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK2-NEXT:    store i64* [[TMP30]], i64** [[TMP39]], align 4
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP40]], align 4
// CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP41]], align 4
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP42]], align 4
// CHECK2-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK2-NEXT:    store i64 [[TMP34]], i64* [[TMP43]], align 8
// CHECK2-NEXT:    [[TMP44:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB4:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l69.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK2-NEXT:    [[TMP45:%.*]] = icmp ne i32 [[TMP44]], 0
// CHECK2-NEXT:    br i1 [[TMP45]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l69(i32 [[TMP1]], i32* [[TMP2]], i32* [[TMP3]]) #[[ATTR2:[0-9]+]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    [[TMP46:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP46]], i32* [[SIZE_CASTED4]], align 4
// CHECK2-NEXT:    [[TMP47:%.*]] = load i32, i32* [[SIZE_CASTED4]], align 4
// CHECK2-NEXT:    [[TMP48:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP49:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP50:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP51:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[TMP51]], i32 0
// CHECK2-NEXT:    [[TMP52:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP53:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP53]], i32 0
// CHECK2-NEXT:    [[TMP54:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    [[TMP55:%.*]] = mul nuw i32 [[TMP54]], 4
// CHECK2-NEXT:    [[TMP56:%.*]] = sext i32 [[TMP55]] to i64
// CHECK2-NEXT:    [[TMP57:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES10]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP57]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.7 to i8*), i32 24, i1 false)
// CHECK2-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i32*
// CHECK2-NEXT:    store i32 [[TMP47]], i32* [[TMP59]], align 4
// CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP61:%.*]] = bitcast i8** [[TMP60]] to i32*
// CHECK2-NEXT:    store i32 [[TMP47]], i32* [[TMP61]], align 4
// CHECK2-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP62]], align 4
// CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32**
// CHECK2-NEXT:    store i32* [[TMP50]], i32** [[TMP64]], align 4
// CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP66:%.*]] = bitcast i8** [[TMP65]] to i32**
// CHECK2-NEXT:    store i32* [[ARRAYIDX5]], i32** [[TMP66]], align 4
// CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP67]], align 4
// CHECK2-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i32**
// CHECK2-NEXT:    store i32* [[TMP52]], i32** [[TMP69]], align 4
// CHECK2-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32**
// CHECK2-NEXT:    store i32* [[ARRAYIDX6]], i32** [[TMP71]], align 4
// CHECK2-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES10]], i32 0, i32 2
// CHECK2-NEXT:    store i64 [[TMP56]], i64* [[TMP72]], align 4
// CHECK2-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP73]], align 4
// CHECK2-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES10]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP77:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP77]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK2-NEXT:    [[TMP78:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK2-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP78]], 0
// CHECK2-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
// CHECK2-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
// CHECK2-NEXT:    store i32 [[SUB16]], i32* [[DOTCAPTURE_EXPR_13]], align 4
// CHECK2-NEXT:    [[TMP79:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_13]], align 4
// CHECK2-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP79]], 1
// CHECK2-NEXT:    [[TMP80:%.*]] = zext i32 [[ADD17]] to i64
// CHECK2-NEXT:    [[KERNEL_ARGS18:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK2-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP81]], align 4
// CHECK2-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 1
// CHECK2-NEXT:    store i32 3, i32* [[TMP82]], align 4
// CHECK2-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP74]], i8*** [[TMP83]], align 4
// CHECK2-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP75]], i8*** [[TMP84]], align 4
// CHECK2-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 4
// CHECK2-NEXT:    store i64* [[TMP76]], i64** [[TMP85]], align 4
// CHECK2-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.8, i32 0, i32 0), i64** [[TMP86]], align 4
// CHECK2-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP87]], align 4
// CHECK2-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP88]], align 4
// CHECK2-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 8
// CHECK2-NEXT:    store i64 [[TMP80]], i64* [[TMP89]], align 8
// CHECK2-NEXT:    [[TMP90:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB4]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l73.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]])
// CHECK2-NEXT:    [[TMP91:%.*]] = icmp ne i32 [[TMP90]], 0
// CHECK2-NEXT:    br i1 [[TMP91]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
// CHECK2:       omp_offload.failed19:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l73(i32 [[TMP47]], i32* [[TMP48]], i32* [[TMP49]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
// CHECK2:       omp_offload.cont20:
// CHECK2-NEXT:    [[TMP92:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP92]], i32* [[SIZE_CASTED21]], align 4
// CHECK2-NEXT:    [[TMP93:%.*]] = load i32, i32* [[SIZE_CASTED21]], align 4
// CHECK2-NEXT:    [[ARRAYIDX22:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP95:%.*]] = bitcast i8** [[TMP94]] to i32*
// CHECK2-NEXT:    store i32 [[TMP93]], i32* [[TMP95]], align 4
// CHECK2-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i32*
// CHECK2-NEXT:    store i32 [[TMP93]], i32* [[TMP97]], align 4
// CHECK2-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP98]], align 4
// CHECK2-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to [10 x i32]**
// CHECK2-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP100]], align 4
// CHECK2-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i32**
// CHECK2-NEXT:    store i32* [[ARRAYIDX22]], i32** [[TMP102]], align 4
// CHECK2-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP103]], align 4
// CHECK2-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
// CHECK2-NEXT:    [[KERNEL_ARGS26:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK2-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP106]], align 4
// CHECK2-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 1
// CHECK2-NEXT:    store i32 2, i32* [[TMP107]], align 4
// CHECK2-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP104]], i8*** [[TMP108]], align 4
// CHECK2-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP105]], i8*** [[TMP109]], align 4
// CHECK2-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 4
// CHECK2-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.11, i32 0, i32 0), i64** [[TMP110]], align 4
// CHECK2-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.12, i32 0, i32 0), i64** [[TMP111]], align 4
// CHECK2-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP112]], align 4
// CHECK2-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP113]], align 4
// CHECK2-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]], i32 0, i32 8
// CHECK2-NEXT:    store i64 0, i64* [[TMP114]], align 8
// CHECK2-NEXT:    [[TMP115:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB4]], i64 -1, i32 1, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l78.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS26]])
// CHECK2-NEXT:    [[TMP116:%.*]] = icmp ne i32 [[TMP115]], 0
// CHECK2-NEXT:    br i1 [[TMP116]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
// CHECK2:       omp_offload.failed27:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l78(i32 [[TMP93]], [10 x i32]* [[A]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
// CHECK2:       omp_offload.cont28:
// CHECK2-NEXT:    [[TMP117:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP117]], i32* [[SIZE_CASTED29]], align 4
// CHECK2-NEXT:    [[TMP118:%.*]] = load i32, i32* [[SIZE_CASTED29]], align 4
// CHECK2-NEXT:    [[ARRAYIDX30:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS31]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32*
// CHECK2-NEXT:    store i32 [[TMP118]], i32* [[TMP120]], align 4
// CHECK2-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS32]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32*
// CHECK2-NEXT:    store i32 [[TMP118]], i32* [[TMP122]], align 4
// CHECK2-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS33]], i32 0, i32 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP123]], align 4
// CHECK2-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS31]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to [10 x i32]**
// CHECK2-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[TMP125]], align 4
// CHECK2-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS32]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP127:%.*]] = bitcast i8** [[TMP126]] to i32**
// CHECK2-NEXT:    store i32* [[ARRAYIDX30]], i32** [[TMP127]], align 4
// CHECK2-NEXT:    [[TMP128:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS33]], i32 0, i32 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP128]], align 4
// CHECK2-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS31]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS32]], i32 0, i32 0
// CHECK2-NEXT:    [[KERNEL_ARGS34:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK2-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 0
// CHECK2-NEXT:    store i32 1, i32* [[TMP131]], align 4
// CHECK2-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 1
// CHECK2-NEXT:    store i32 2, i32* [[TMP132]], align 4
// CHECK2-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 2
// CHECK2-NEXT:    store i8** [[TMP129]], i8*** [[TMP133]], align 4
// CHECK2-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 3
// CHECK2-NEXT:    store i8** [[TMP130]], i8*** [[TMP134]], align 4
// CHECK2-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 4
// CHECK2-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.15, i32 0, i32 0), i64** [[TMP135]], align 4
// CHECK2-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 5
// CHECK2-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.16, i32 0, i32 0), i64** [[TMP136]], align 4
// CHECK2-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 6
// CHECK2-NEXT:    store i8** null, i8*** [[TMP137]], align 4
// CHECK2-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 7
// CHECK2-NEXT:    store i8** null, i8*** [[TMP138]], align 4
// CHECK2-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 8
// CHECK2-NEXT:    store i64 0, i64* [[TMP139]], align 8
// CHECK2-NEXT:    [[TMP140:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB4]], i64 -1, i32 1, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l81.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]])
// CHECK2-NEXT:    [[TMP141:%.*]] = icmp ne i32 [[TMP140]], 0
// CHECK2-NEXT:    br i1 [[TMP141]], label [[OMP_OFFLOAD_FAILED35:%.*]], label [[OMP_OFFLOAD_CONT36:%.*]]
// CHECK2:       omp_offload.failed35:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l81(i32 [[TMP118]], [10 x i32]* [[A]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT36]]
// CHECK2:       omp_offload.cont36:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l69
// CHECK2-SAME: (i32 noundef [[SIZE:%.*]], i32* noundef [[OUTPUT:%.*]], i32* noundef [[INPUT:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[INPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[SIZE_CASTED:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[OUTPUT]], i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[INPUT]], i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB4]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP1]], i32* [[TMP2]], i32* [[TMP3]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[SIZE:%.*]], i32* noundef [[OUTPUT:%.*]], i32* noundef [[INPUT:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[INPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[OUTPUT1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I5:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SIZE_CASTED:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[OUTPUT]], i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[INPUT]], i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i32 0
// CHECK2-NEXT:    store i32 0, i32* [[OUTPUT1]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = ptrtoint i32* [[TMP1]] to i64
// CHECK2-NEXT:    [[TMP3:%.*]] = ptrtoint i32* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP4:%.*]] = sub i64 [[TMP2]], [[TMP3]]
// CHECK2-NEXT:    [[TMP5:%.*]] = sdiv exact i64 [[TMP4]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[OUTPUT1]], i64 [[TMP5]]
// CHECK2-NEXT:    store i32* [[TMP6]], i32** [[TMP]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP8]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB4]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP22]], i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[TMP]], align 4
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP20]], i32 [[TMP21]], i32 [[TMP23]], i32* [[TMP24]], i32* [[TMP25]])
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP29]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP31:%.*]] = bitcast i32* [[OUTPUT1]] to i8*
// CHECK2-NEXT:    store i8* [[TMP31]], i8** [[TMP30]], align 4
// CHECK2-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK2-NEXT:    [[TMP34:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP35:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP33]], i32 1, i32 4, i8* [[TMP34]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP35]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[OUTPUT1]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP33]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[OUTPUT1]], align 4
// CHECK2-NEXT:    [[TMP39:%.*]] = atomicrmw add i32* [[ARRAYIDX]], i32 [[TMP38]] monotonic, align 4
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[SIZE:%.*]], i32* noundef [[OUTPUT:%.*]], i32* noundef [[INPUT:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[INPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP4:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[I5:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[OUTPUT]], i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[INPUT]], i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i32 0
// CHECK2-NEXT:    store i32 0, i32* [[OUTPUT3]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint i32* [[TMP7]] to i64
// CHECK2-NEXT:    [[TMP9:%.*]] = ptrtoint i32* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[OUTPUT3]], i64 [[TMP11]]
// CHECK2-NEXT:    store i32* [[TMP12]], i32** [[_TMP4]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP14]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP15]], [[TMP16]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP17]], [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I5]], align 4
// CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i32 [[TMP24]]
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[ARRAYIDX8]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[_TMP4]], align 4
// CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i32 0
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[ARRAYIDX9]], align 4
// CHECK2-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP25]]
// CHECK2-NEXT:    store i32 [[ADD10]], i32* [[ARRAYIDX9]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK2-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast i32* [[OUTPUT3]] to i8*
// CHECK2-NEXT:    store i8* [[TMP32]], i8** [[TMP31]], align 4
// CHECK2-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
// CHECK2-NEXT:    [[TMP35:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP34]], i32 1, i32 4, i8* [[TMP35]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP36]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[OUTPUT3]], align 4
// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP37]], [[TMP38]]
// CHECK2-NEXT:    store i32 [[ADD12]], i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP34]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[OUTPUT3]], align 4
// CHECK2-NEXT:    [[TMP40:%.*]] = atomicrmw add i32* [[ARRAYIDX]], i32 [[TMP39]] monotonic, align 4
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
// CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2
// CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l73
// CHECK2-SAME: (i32 noundef [[SIZE:%.*]], i32* noundef [[OUTPUT:%.*]], i32* noundef [[INPUT:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[INPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[SIZE_CASTED:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[OUTPUT]], i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[INPUT]], i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB4]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32* [[TMP2]], i32* [[TMP3]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[SIZE:%.*]], i32* noundef [[OUTPUT:%.*]], i32* noundef [[INPUT:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[INPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[OUTPUT2:%.*]] = alloca [3 x i32], align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SIZE_CASTED:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[OUTPUT]], i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[INPUT]], i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i32 0
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i32 2
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [3 x i32], [3 x i32]* [[OUTPUT2]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i32 3
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP2]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP2]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = ptrtoint i32* [[TMP3]] to i64
// CHECK2-NEXT:    [[TMP5:%.*]] = ptrtoint i32* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP6:%.*]] = sub i64 [[TMP4]], [[TMP5]]
// CHECK2-NEXT:    [[TMP7:%.*]] = sdiv exact i64 [[TMP6]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast [3 x i32]* [[OUTPUT2]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr i32, i32* [[TMP8]], i64 [[TMP7]]
// CHECK2-NEXT:    store i32* [[TMP9]], i32** [[TMP]], align 4
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [3 x i32]* [[OUTPUT2]] to i32*
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB5:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP12]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP16]], [[TMP17]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP18]], [[COND_TRUE]] ], [ [[TMP19]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP20]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP21]], [[TMP22]]
// CHECK2-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP25]], i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP]], align 4
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP23]], i32 [[TMP24]], i32 [[TMP26]], i32* [[TMP27]], i32* [[TMP28]])
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP29]], [[TMP30]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP32]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP34:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP34]], i8** [[TMP33]], align 4
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP38:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], i32 1, i32 4, i8* [[TMP37]], void (i8*, i8*)* @.omp.reduction.reduction_func.6, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP38]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr i32, i32* [[ARRAYIDX]], i32 3
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[ARRAYIDX]], [[TMP39]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE13:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST9:%.*]] = phi i32* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST9]], align 4
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP40]], [[TMP41]]
// CHECK2-NEXT:    store i32 [[ADD10]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST9]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST9]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE12:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP39]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE12]], label [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done13:
// CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr i32, i32* [[ARRAYIDX]], i32 3
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY14:%.*]] = icmp eq i32* [[ARRAYIDX]], [[TMP42]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY14]], label [[OMP_ARRAYCPY_DONE21:%.*]], label [[OMP_ARRAYCPY_BODY15:%.*]]
// CHECK2:       omp.arraycpy.body15:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST16:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT19:%.*]], [[OMP_ARRAYCPY_BODY15]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST17:%.*]] = phi i32* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY15]] ]
// CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST16]], align 4
// CHECK2-NEXT:    [[TMP44:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST17]], i32 [[TMP43]] monotonic, align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT18]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST17]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT19]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST16]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT18]], [[TMP42]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_BODY15]]
// CHECK2:       omp.arraycpy.done21:
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[SIZE:%.*]], i32* noundef [[OUTPUT:%.*]], i32* noundef [[INPUT:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[INPUT_ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[OUTPUT4:%.*]] = alloca [3 x i32], align 4
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[OUTPUT]], i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    store i32* [[INPUT]], i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP6]], i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, i32* [[TMP7]], i32 2
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [3 x i32], [3 x i32]* [[OUTPUT4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i32 3
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP8]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[OMP_PRECOND_THEN]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[OUTPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = ptrtoint i32* [[TMP9]] to i64
// CHECK2-NEXT:    [[TMP11:%.*]] = ptrtoint i32* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP12:%.*]] = sub i64 [[TMP10]], [[TMP11]]
// CHECK2-NEXT:    [[TMP13:%.*]] = sdiv exact i64 [[TMP12]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast [3 x i32]* [[OUTPUT4]] to i32*
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr i32, i32* [[TMP14]], i64 [[TMP13]]
// CHECK2-NEXT:    store i32* [[TMP15]], i32** [[_TMP5]], align 4
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [3 x i32]* [[OUTPUT4]] to i32*
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP17]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
// CHECK2-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[INPUT_ADDR]], align 4
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4
// CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i32 [[TMP27]]
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX9]], align 4
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[_TMP5]], align 4
// CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i32 0
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[ARRAYIDX10]], align 4
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP30]], [[TMP28]]
// CHECK2-NEXT:    store i32 [[ADD11]], i32* [[ARRAYIDX10]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP31]], 1
// CHECK2-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP35:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP35]], i8** [[TMP34]], align 4
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP37]], i32 1, i32 4, i8* [[TMP38]], void (i8*, i8*)* @.omp.reduction.reduction_func.5, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP39]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr i32, i32* [[ARRAYIDX]], i32 3
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[ARRAYIDX]], [[TMP40]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST13:%.*]] = phi i32* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST13]], align 4
// CHECK2-NEXT:    [[TMP42:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP41]], [[TMP42]]
// CHECK2-NEXT:    store i32 [[ADD14]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST13]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST13]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP40]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done17:
// CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP37]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP43:%.*]] = getelementptr i32, i32* [[ARRAYIDX]], i32 3
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY18:%.*]] = icmp eq i32* [[ARRAYIDX]], [[TMP43]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY18]], label [[OMP_ARRAYCPY_DONE25:%.*]], label [[OMP_ARRAYCPY_BODY19:%.*]]
// CHECK2:       omp.arraycpy.body19:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST20:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT23:%.*]], [[OMP_ARRAYCPY_BODY19]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST21:%.*]] = phi i32* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT22:%.*]], [[OMP_ARRAYCPY_BODY19]] ]
// CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 4
// CHECK2-NEXT:    [[TMP45:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 [[TMP44]] monotonic, align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT22]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT23]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE24:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT22]], [[TMP43]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE24]], label [[OMP_ARRAYCPY_DONE25]], label [[OMP_ARRAYCPY_BODY19]]
// CHECK2:       omp.arraycpy.done25:
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.5
// CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i32 3
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.6
// CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i32 3
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l78
// CHECK2-SAME: (i32 noundef [[SIZE:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR6:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK2-NEXT:    [[SIZE_CASTED:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP2]], [10 x i32]* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[SIZE:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK2-NEXT:    [[A2:%.*]] = alloca [2 x i32], align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 1
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[A2]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i32 2
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP1]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP1]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP2:%.*]] = bitcast [10 x i32]* [[TMP0]] to i32*
// CHECK2-NEXT:    [[TMP3:%.*]] = ptrtoint i32* [[TMP2]] to i64
// CHECK2-NEXT:    [[TMP4:%.*]] = ptrtoint i32* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
// CHECK2-NEXT:    [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP7:%.*]] = bitcast [2 x i32]* [[A2]] to i32*
// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP7]], i64 [[TMP6]]
// CHECK2-NEXT:    [[TMP9:%.*]] = bitcast i32* [[TMP8]] to [10 x i32]*
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x i32]* [[A2]] to i32*
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP10]], [[TMP11]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP16]], i32 1, i32 4, i8* [[TMP17]], void (i8*, i8*)* @.omp.reduction.reduction_func.10, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP18]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr i32, i32* [[ARRAYIDX]], i32 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[ARRAYIDX]], [[TMP19]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP19]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done6:
// CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP16]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr i32, i32* [[ARRAYIDX]], i32 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[ARRAYIDX]], [[TMP22]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
// CHECK2:       omp.arraycpy.body8:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP23]] monotonic, align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP22]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
// CHECK2:       omp.arraycpy.done14:
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.10
// CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i32 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3sumPiiS__l81
// CHECK2-SAME: (i32 noundef [[SIZE:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK2-NEXT:    [[SIZE_CASTED:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[SIZE_CASTED]], align 4
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x i32]*)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32 [[TMP2]], [10 x i32]* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..13
// CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[SIZE:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[SIZE_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK2-NEXT:    [[A1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK2-NEXT:    store i32 [[SIZE]], i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 3
// CHECK2-NEXT:    store i32 0, i32* [[A1]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = bitcast [10 x i32]* [[TMP0]] to i32*
// CHECK2-NEXT:    [[TMP2:%.*]] = ptrtoint i32* [[TMP1]] to i64
// CHECK2-NEXT:    [[TMP3:%.*]] = ptrtoint i32* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP4:%.*]] = sub i64 [[TMP2]], [[TMP3]]
// CHECK2-NEXT:    [[TMP5:%.*]] = sdiv exact i64 [[TMP4]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[A1]], i64 [[TMP5]]
// CHECK2-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to [10 x i32]*
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[SIZE_ADDR]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i32* [[A1]] to i8*
// CHECK2-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32 1, i32 4, i8* [[TMP15]], void (i8*, i8*)* @.omp.reduction.reduction_func.14, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP16]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[A1]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A1]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = atomicrmw add i32* [[ARRAYIDX]], i32 [[TMP19]] monotonic, align 4
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.14
// CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@main
// CHECK2-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SIZE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[ARRAY:%.*]] = alloca i32*, align 4
// CHECK2-NEXT:    [[RESULT:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK2-NEXT:    store i32 100, i32* [[SIZE]], align 4
// CHECK2-NEXT:    [[CALL:%.*]] = call noalias noundef nonnull i8* @_Znaj(i32 noundef 400) #[[ATTR10:[0-9]+]]
// CHECK2-NEXT:    [[TMP0:%.*]] = bitcast i8* [[CALL]] to i32*
// CHECK2-NEXT:    store i32* [[TMP0]], i32** [[ARRAY]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[RESULT]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARRAY]], align 4
// CHECK2-NEXT:    call void @_Z3sumPiiS_(i32* noundef [[TMP1]], i32 noundef 100, i32* noundef [[RESULT]])
// CHECK2-NEXT:    ret i32 0
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK2-SAME: () #[[ATTR9:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK2-NEXT:    ret void
//