// 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 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s
// expected-no-diagnostics

#define N 1000

template<typename T>
void compute_min_max() {
  T min_t = 1000;
  T max_t = 0;
  T *arr_t = new T[N];
  for (int i = 0; i < N; i++)
    arr_t[i] = i;
#pragma omp target data map(to : arr_t[0 : N])
  {
#pragma omp target teams distribute parallel for reduction(min : min_t)
    for (int j = 0; j < N; j = j + 1)
      min_t = __builtin_fmin(min_t, arr_t[j]);

#pragma omp target teams distribute parallel for reduction(max : max_t)
    for (int j = 0; j < N; j = j + 1)
      max_t = __builtin_fmax(max_t, arr_t[j]);
  }
  delete[] arr_t;
}

int main()
{
  compute_min_max<short>();
  compute_min_max<unsigned short>();
  compute_min_max<int>();
  compute_min_max<unsigned int>();
  compute_min_max<long long>();
  compute_min_max<unsigned long long>();
  compute_min_max<float>();
  compute_min_max<double>();
}

// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIsEvv_l17
// CHECK-SAME: (ptr noundef nonnull align 2 dereferenceable(2) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20:![0-9]+]], !align [[META21:![0-9]+]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    store i16 32767, ptr addrspace(5) [[TMP3]], align 2
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
// CHECK-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP18]] to double
// CHECK-NEXT:    [[TMP19:%.*]] = load i16, ptr addrspace(5) [[TMP3]], align 2
// CHECK-NEXT:    [[TMP20:%.*]] = fptosi double [[CONV]] to i16
// CHECK-NEXT:    [[XTEAM_MIN:%.*]] = call i16 @llvm.smin.i16(i16 [[TMP19]], i16 [[TMP20]])
// CHECK-NEXT:    store i16 [[XTEAM_MIN]], ptr addrspace(5) [[TMP3]], align 2
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP22:%.*]] = mul i32 [[TMP21]], 1
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP27:%.*]] = load i16, ptr addrspace(5) [[TMP3]], align 2
// CHECK-NEXT:    call void @__kmpc_xteamr_s_16x64(i16 [[TMP27]], ptr [[TMP2]], ptr [[TMP25]], ptr [[TMP26]], ptr @__kmpc_rfun_min_s, ptr @__kmpc_rfun_min_lds_s, i16 32767, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIsEvv_l21
// CHECK-SAME: (ptr noundef nonnull align 2 dereferenceable(2) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META21]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    store i16 -32768, ptr addrspace(5) [[TMP3]], align 2
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
// CHECK-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP18]] to double
// CHECK-NEXT:    [[TMP19:%.*]] = load i16, ptr addrspace(5) [[TMP3]], align 2
// CHECK-NEXT:    [[TMP20:%.*]] = fptosi double [[CONV]] to i16
// CHECK-NEXT:    [[XTEAM_MAX:%.*]] = call i16 @llvm.smax.i16(i16 [[TMP19]], i16 [[TMP20]])
// CHECK-NEXT:    store i16 [[XTEAM_MAX]], ptr addrspace(5) [[TMP3]], align 2
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP22:%.*]] = mul i32 [[TMP21]], 1
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP27:%.*]] = load i16, ptr addrspace(5) [[TMP3]], align 2
// CHECK-NEXT:    call void @__kmpc_xteamr_s_16x64(i16 [[TMP27]], ptr [[TMP2]], ptr [[TMP25]], ptr [[TMP26]], ptr @__kmpc_rfun_max_s, ptr @__kmpc_rfun_max_lds_s, i16 -32768, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17
// CHECK-SAME: (ptr noundef nonnull align 2 dereferenceable(2) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META21]]
// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_kernel_environment to ptr), ptr [[DYN_PTR]])
// CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
// CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK:       user_code.entry:
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1:[0-9]+]] to ptr))
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP2]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
// CHECK-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_omp_outlined(ptr [[TMP4]], ptr [[DOTZERO_ADDR_ASCAST]], ptr [[TMP0]], ptr [[TMP3]]) #[[ATTR1:[0-9]+]]
// CHECK-NEXT:    call void @__kmpc_target_deinit()
// CHECK-NEXT:    ret void
// CHECK:       worker.exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 2 dereferenceable(2) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T1]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META21]]
// CHECK-NEXT:    store i16 -1, ptr [[MIN_T1_ASCAST]], align 2
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2:[0-9]+]] to ptr), i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 999
// CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 1000
// CHECK-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP8]] to ptr
// CHECK-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
// CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP10]] to ptr
// CHECK-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
// CHECK-NEXT:    store ptr [[MIN_T1_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
// CHECK-NEXT:    store ptr [[TMP11]], ptr [[TMP17]], align 8
// CHECK-NEXT:    call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP2]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 4, i32 0)
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP24]], 999
// CHECK-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK:       cond.true6:
// CHECK-NEXT:    br label [[COND_END8:%.*]]
// CHECK:       cond.false7:
// CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END8]]
// CHECK:       cond.end8:
// CHECK-NEXT:    [[COND9:%.*]] = phi i32 [ 999, [[COND_TRUE6]] ], [ [[TMP25]], [[COND_FALSE7]] ]
// CHECK-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP26]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]])
// CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MIN_T1_ASCAST]], ptr [[TMP27]], align 8
// CHECK-NEXT:    %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
// CHECK-NEXT:    [[TMP28:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 2, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.1, ptr @_omp_reduction_inter_warp_copy_func.2, ptr @_omp_reduction_list_to_global_copy_func, ptr @_omp_reduction_list_to_global_reduce_func, ptr @_omp_reduction_global_to_list_copy_func, ptr @_omp_reduction_global_to_list_reduce_func)
// CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP28]], 1
// CHECK-NEXT:    br i1 [[TMP29]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP30:%.*]] = load i16, ptr [[TMP0]], align 2
// CHECK-NEXT:    [[CONV:%.*]] = zext i16 [[TMP30]] to i32
// CHECK-NEXT:    [[TMP31:%.*]] = load i16, ptr [[MIN_T1_ASCAST]], align 2
// CHECK-NEXT:    [[CONV10:%.*]] = zext i16 [[TMP31]] to i32
// CHECK-NEXT:    [[CMP11:%.*]] = icmp slt i32 [[CONV]], [[CONV10]]
// CHECK-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK:       cond.true12:
// CHECK-NEXT:    [[TMP32:%.*]] = load i16, ptr [[TMP0]], align 2
// CHECK-NEXT:    br label [[COND_END14:%.*]]
// CHECK:       cond.false13:
// CHECK-NEXT:    [[TMP33:%.*]] = load i16, ptr [[MIN_T1_ASCAST]], align 2
// CHECK-NEXT:    br label [[COND_END14]]
// CHECK:       cond.end14:
// CHECK-NEXT:    [[COND15:%.*]] = phi i16 [ [[TMP32]], [[COND_TRUE12]] ], [ [[TMP33]], [[COND_FALSE13]] ]
// CHECK-NEXT:    store i16 [[COND15]], ptr [[TMP0]], align 2
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_omp_outlined_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 2 dereferenceable(2) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[MIN_T2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T2]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META21]]
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    store i16 -1, ptr [[MIN_T2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3:[0-9]+]] to ptr), i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
// CHECK-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP9:%.*]] = load i16, ptr [[MIN_T2_ASCAST]], align 2
// CHECK-NEXT:    [[CONV4:%.*]] = uitofp i16 [[TMP9]] to double
// CHECK-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[TMP10]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
// CHECK-NEXT:    [[CONV5:%.*]] = uitofp i16 [[TMP12]] to double
// CHECK-NEXT:    [[TMP13:%.*]] = call nsz double @llvm.minnum.f64(double [[CONV4]], double [[CONV5]])
// CHECK-NEXT:    [[CONV6:%.*]] = fptoui double [[TMP13]] to i16
// CHECK-NEXT:    store i16 [[CONV6]], ptr [[MIN_T2_ASCAST]], align 2
// CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK:       omp.body.continue:
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]])
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MIN_T2_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 2, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func, ptr @_omp_reduction_inter_warp_copy_func)
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[TMP17]], 1
// CHECK-NEXT:    br i1 [[TMP18]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP19:%.*]] = load i16, ptr [[TMP0]], align 2
// CHECK-NEXT:    [[CONV8:%.*]] = zext i16 [[TMP19]] to i32
// CHECK-NEXT:    [[TMP20:%.*]] = load i16, ptr [[MIN_T2_ASCAST]], align 2
// CHECK-NEXT:    [[CONV9:%.*]] = zext i16 [[TMP20]] to i32
// CHECK-NEXT:    [[CMP10:%.*]] = icmp slt i32 [[CONV8]], [[CONV9]]
// CHECK-NEXT:    br i1 [[CMP10]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    [[TMP21:%.*]] = load i16, ptr [[TMP0]], align 2
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP22:%.*]] = load i16, ptr [[MIN_T2_ASCAST]], align 2
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i16 [ [[TMP21]], [[COND_TRUE]] ], [ [[TMP22]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i16 [[COND]], ptr [[TMP0]], align 2
// 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: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6:[0-9]+]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i16, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i16, ptr [[TMP9]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = sext i16 [[TMP12]] to i32
// CHECK-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP15:%.*]] = trunc i32 [[TMP14]] to i16
// CHECK-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP13]], i16 [[TMP6]], i16 [[TMP15]])
// CHECK-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i16
// CHECK-NEXT:    store i16 [[TMP17]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 2
// CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i16, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i16, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP22:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP23:%.*]] = and i1 [[TMP21]], [[TMP22]]
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP25:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i16 [[TMP25]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP24]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP29:%.*]] = and i1 [[TMP27]], [[TMP28]]
// CHECK-NEXT:    [[TMP30:%.*]] = or i1 [[TMP20]], [[TMP23]]
// CHECK-NEXT:    [[TMP31:%.*]] = or i1 [[TMP30]], [[TMP29]]
// CHECK-NEXT:    br i1 [[TMP31]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP33:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP34:%.*]] = and i1 [[TMP32]], [[TMP33]]
// CHECK-NEXT:    br i1 [[TMP34]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP38:%.*]] = load ptr, ptr [[TMP37]], align 8
// CHECK-NEXT:    [[TMP39:%.*]] = load i16, ptr [[TMP36]], align 2
// CHECK-NEXT:    store i16 [[TMP39]], ptr [[TMP38]], align 2
// 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: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4:[0-9]+]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP9:%.*]] = load i16, ptr [[TMP7]], align 2
// CHECK-NEXT:    store volatile i16 [[TMP9]], ptr addrspace(3) [[TMP8]], align 2
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP10]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load volatile i16, ptr addrspace(3) [[TMP11]], align 2
// CHECK-NEXT:    store i16 [[TMP14]], ptr [[TMP13]], align 2
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.1
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i16, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i16, ptr [[TMP9]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = sext i16 [[TMP12]] to i32
// CHECK-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP15:%.*]] = trunc i32 [[TMP14]] to i16
// CHECK-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP13]], i16 [[TMP6]], i16 [[TMP15]])
// CHECK-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i16
// CHECK-NEXT:    store i16 [[TMP17]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 2
// CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i16, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i16, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP22:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP23:%.*]] = and i1 [[TMP21]], [[TMP22]]
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP25:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i16 [[TMP25]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP24]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP29:%.*]] = and i1 [[TMP27]], [[TMP28]]
// CHECK-NEXT:    [[TMP30:%.*]] = or i1 [[TMP20]], [[TMP23]]
// CHECK-NEXT:    [[TMP31:%.*]] = or i1 [[TMP30]], [[TMP29]]
// CHECK-NEXT:    br i1 [[TMP31]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP33:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP34:%.*]] = and i1 [[TMP32]], [[TMP33]]
// CHECK-NEXT:    br i1 [[TMP34]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP38:%.*]] = load ptr, ptr [[TMP37]], align 8
// CHECK-NEXT:    [[TMP39:%.*]] = load i16, ptr [[TMP36]], align 2
// CHECK-NEXT:    store i16 [[TMP39]], ptr [[TMP38]], align 2
// 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.2
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP9:%.*]] = load i16, ptr [[TMP7]], align 2
// CHECK-NEXT:    store volatile i16 [[TMP9]], ptr addrspace(3) [[TMP8]], align 2
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP10]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load volatile i16, ptr addrspace(3) [[TMP11]], align 2
// CHECK-NEXT:    store i16 [[TMP14]], ptr [[TMP13]], align 2
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i16, ptr [[TMP7]], align 2
// CHECK-NEXT:    store i16 [[TMP10]], ptr [[TMP9]], align 2
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i16, ptr [[TMP9]], align 2
// CHECK-NEXT:    store i16 [[TMP10]], ptr [[TMP7]], align 2
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21
// CHECK-SAME: (ptr noundef nonnull align 2 dereferenceable(2) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR4]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META21]]
// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_kernel_environment to ptr), ptr [[DYN_PTR]])
// CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
// CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK:       user_code.entry:
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP2]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
// CHECK-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_omp_outlined(ptr [[TMP4]], ptr [[DOTZERO_ADDR_ASCAST]], ptr [[TMP0]], ptr [[TMP3]]) #[[ATTR1]]
// CHECK-NEXT:    call void @__kmpc_target_deinit()
// CHECK-NEXT:    ret void
// CHECK:       worker.exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 2 dereferenceable(2) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T1]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META21]]
// CHECK-NEXT:    store i16 0, ptr [[MAX_T1_ASCAST]], align 2
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 999
// CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 1000
// CHECK-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP8]] to ptr
// CHECK-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
// CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP10]] to ptr
// CHECK-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
// CHECK-NEXT:    store ptr [[MAX_T1_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
// CHECK-NEXT:    store ptr [[TMP11]], ptr [[TMP17]], align 8
// CHECK-NEXT:    call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP2]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 4, i32 0)
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP24]], 999
// CHECK-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK:       cond.true6:
// CHECK-NEXT:    br label [[COND_END8:%.*]]
// CHECK:       cond.false7:
// CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END8]]
// CHECK:       cond.end8:
// CHECK-NEXT:    [[COND9:%.*]] = phi i32 [ 999, [[COND_TRUE6]] ], [ [[TMP25]], [[COND_FALSE7]] ]
// CHECK-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP26]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]])
// CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MAX_T1_ASCAST]], ptr [[TMP27]], align 8
// CHECK-NEXT:    %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
// CHECK-NEXT:    [[TMP28:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 2, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.5, ptr @_omp_reduction_inter_warp_copy_func.6, ptr @_omp_reduction_list_to_global_copy_func.7, ptr @_omp_reduction_list_to_global_reduce_func.8, ptr @_omp_reduction_global_to_list_copy_func.9, ptr @_omp_reduction_global_to_list_reduce_func.10)
// CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP28]], 1
// CHECK-NEXT:    br i1 [[TMP29]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP30:%.*]] = load i16, ptr [[TMP0]], align 2
// CHECK-NEXT:    [[CONV:%.*]] = zext i16 [[TMP30]] to i32
// CHECK-NEXT:    [[TMP31:%.*]] = load i16, ptr [[MAX_T1_ASCAST]], align 2
// CHECK-NEXT:    [[CONV10:%.*]] = zext i16 [[TMP31]] to i32
// CHECK-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[CONV]], [[CONV10]]
// CHECK-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
// CHECK:       cond.true12:
// CHECK-NEXT:    [[TMP32:%.*]] = load i16, ptr [[TMP0]], align 2
// CHECK-NEXT:    br label [[COND_END14:%.*]]
// CHECK:       cond.false13:
// CHECK-NEXT:    [[TMP33:%.*]] = load i16, ptr [[MAX_T1_ASCAST]], align 2
// CHECK-NEXT:    br label [[COND_END14]]
// CHECK:       cond.end14:
// CHECK-NEXT:    [[COND15:%.*]] = phi i16 [ [[TMP32]], [[COND_TRUE12]] ], [ [[TMP33]], [[COND_FALSE13]] ]
// CHECK-NEXT:    store i16 [[COND15]], ptr [[TMP0]], align 2
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_omp_outlined_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 2 dereferenceable(2) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[MAX_T2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T2]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META21]]
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    store i16 0, ptr [[MAX_T2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
// CHECK-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP9:%.*]] = load i16, ptr [[MAX_T2_ASCAST]], align 2
// CHECK-NEXT:    [[CONV4:%.*]] = uitofp i16 [[TMP9]] to double
// CHECK-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[TMP10]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
// CHECK-NEXT:    [[CONV5:%.*]] = uitofp i16 [[TMP12]] to double
// CHECK-NEXT:    [[TMP13:%.*]] = call nsz double @llvm.maxnum.f64(double [[CONV4]], double [[CONV5]])
// CHECK-NEXT:    [[CONV6:%.*]] = fptoui double [[TMP13]] to i16
// CHECK-NEXT:    store i16 [[CONV6]], ptr [[MAX_T2_ASCAST]], align 2
// CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK:       omp.body.continue:
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]])
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MAX_T2_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 2, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.3, ptr @_omp_reduction_inter_warp_copy_func.4)
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[TMP17]], 1
// CHECK-NEXT:    br i1 [[TMP18]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP19:%.*]] = load i16, ptr [[TMP0]], align 2
// CHECK-NEXT:    [[CONV8:%.*]] = zext i16 [[TMP19]] to i32
// CHECK-NEXT:    [[TMP20:%.*]] = load i16, ptr [[MAX_T2_ASCAST]], align 2
// CHECK-NEXT:    [[CONV9:%.*]] = zext i16 [[TMP20]] to i32
// CHECK-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[CONV8]], [[CONV9]]
// CHECK-NEXT:    br i1 [[CMP10]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    [[TMP21:%.*]] = load i16, ptr [[TMP0]], align 2
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP22:%.*]] = load i16, ptr [[MAX_T2_ASCAST]], align 2
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i16 [ [[TMP21]], [[COND_TRUE]] ], [ [[TMP22]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i16 [[COND]], ptr [[TMP0]], align 2
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.3
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i16, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i16, ptr [[TMP9]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = sext i16 [[TMP12]] to i32
// CHECK-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP15:%.*]] = trunc i32 [[TMP14]] to i16
// CHECK-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP13]], i16 [[TMP6]], i16 [[TMP15]])
// CHECK-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i16
// CHECK-NEXT:    store i16 [[TMP17]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 2
// CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i16, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i16, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP22:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP23:%.*]] = and i1 [[TMP21]], [[TMP22]]
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP25:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i16 [[TMP25]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP24]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP29:%.*]] = and i1 [[TMP27]], [[TMP28]]
// CHECK-NEXT:    [[TMP30:%.*]] = or i1 [[TMP20]], [[TMP23]]
// CHECK-NEXT:    [[TMP31:%.*]] = or i1 [[TMP30]], [[TMP29]]
// CHECK-NEXT:    br i1 [[TMP31]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP33:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP34:%.*]] = and i1 [[TMP32]], [[TMP33]]
// CHECK-NEXT:    br i1 [[TMP34]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP38:%.*]] = load ptr, ptr [[TMP37]], align 8
// CHECK-NEXT:    [[TMP39:%.*]] = load i16, ptr [[TMP36]], align 2
// CHECK-NEXT:    store i16 [[TMP39]], ptr [[TMP38]], align 2
// 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.4
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP9:%.*]] = load i16, ptr [[TMP7]], align 2
// CHECK-NEXT:    store volatile i16 [[TMP9]], ptr addrspace(3) [[TMP8]], align 2
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP10]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load volatile i16, ptr addrspace(3) [[TMP11]], align 2
// CHECK-NEXT:    store i16 [[TMP14]], ptr [[TMP13]], align 2
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.5
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i16, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i16, ptr [[TMP9]], align 2
// CHECK-NEXT:    [[TMP13:%.*]] = sext i16 [[TMP12]] to i32
// CHECK-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP15:%.*]] = trunc i32 [[TMP14]] to i16
// CHECK-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP13]], i16 [[TMP6]], i16 [[TMP15]])
// CHECK-NEXT:    [[TMP17:%.*]] = trunc i32 [[TMP16]] to i16
// CHECK-NEXT:    store i16 [[TMP17]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 2
// CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i16, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i16, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP22:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP23:%.*]] = and i1 [[TMP21]], [[TMP22]]
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP25:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i16 [[TMP25]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP24]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP29:%.*]] = and i1 [[TMP27]], [[TMP28]]
// CHECK-NEXT:    [[TMP30:%.*]] = or i1 [[TMP20]], [[TMP23]]
// CHECK-NEXT:    [[TMP31:%.*]] = or i1 [[TMP30]], [[TMP29]]
// CHECK-NEXT:    br i1 [[TMP31]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP33:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP34:%.*]] = and i1 [[TMP32]], [[TMP33]]
// CHECK-NEXT:    br i1 [[TMP34]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP38:%.*]] = load ptr, ptr [[TMP37]], align 8
// CHECK-NEXT:    [[TMP39:%.*]] = load i16, ptr [[TMP36]], align 2
// CHECK-NEXT:    store i16 [[TMP39]], ptr [[TMP38]], align 2
// 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.6
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP9:%.*]] = load i16, ptr [[TMP7]], align 2
// CHECK-NEXT:    store volatile i16 [[TMP9]], ptr addrspace(3) [[TMP8]], align 2
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP10]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load volatile i16, ptr addrspace(3) [[TMP11]], align 2
// CHECK-NEXT:    store i16 [[TMP14]], ptr [[TMP13]], align 2
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.7
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i16, ptr [[TMP7]], align 2
// CHECK-NEXT:    store i16 [[TMP10]], ptr [[TMP9]], align 2
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.8
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.9
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i16, ptr [[TMP9]], align 2
// CHECK-NEXT:    store i16 [[TMP10]], ptr [[TMP7]], align 2
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.10
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxItEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIiEvv_l17
// CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25:![0-9]+]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    store i32 2147483647, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP18]] to double
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = fptosi double [[CONV]] to i32
// CHECK-NEXT:    [[XTEAM_MIN:%.*]] = call i32 @llvm.smin.i32(i32 [[TMP19]], i32 [[TMP20]])
// CHECK-NEXT:    store i32 [[XTEAM_MIN]], ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP22:%.*]] = mul i32 [[TMP21]], 1
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP27:%.*]] = load i32, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_xteamr_i_16x64(i32 [[TMP27]], ptr [[TMP2]], ptr [[TMP25]], ptr [[TMP26]], ptr @__kmpc_rfun_min_i, ptr @__kmpc_rfun_min_lds_i, i32 2147483647, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIiEvv_l21
// CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    store i32 -2147483648, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP18]] to double
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = fptosi double [[CONV]] to i32
// CHECK-NEXT:    [[XTEAM_MAX:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP19]], i32 [[TMP20]])
// CHECK-NEXT:    store i32 [[XTEAM_MAX]], ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP22:%.*]] = mul i32 [[TMP21]], 1
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP27:%.*]] = load i32, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_xteamr_i_16x64(i32 [[TMP27]], ptr [[TMP2]], ptr [[TMP25]], ptr [[TMP26]], ptr @__kmpc_rfun_max_i, ptr @__kmpc_rfun_max_lds_i, i32 -2147483648, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17
// CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR4]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_kernel_environment to ptr), ptr [[DYN_PTR]])
// CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
// CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK:       user_code.entry:
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP2]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
// CHECK-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_omp_outlined(ptr [[TMP4]], ptr [[DOTZERO_ADDR_ASCAST]], ptr [[TMP0]], ptr [[TMP3]]) #[[ATTR1]]
// CHECK-NEXT:    call void @__kmpc_target_deinit()
// CHECK-NEXT:    ret void
// CHECK:       worker.exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T1]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    store i32 -1, ptr [[MIN_T1_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 999
// CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 1000
// CHECK-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP8]] to ptr
// CHECK-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
// CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP10]] to ptr
// CHECK-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
// CHECK-NEXT:    store ptr [[MIN_T1_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
// CHECK-NEXT:    store ptr [[TMP11]], ptr [[TMP17]], align 8
// CHECK-NEXT:    call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP2]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 4, i32 0)
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP24]], 999
// CHECK-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK:       cond.true6:
// CHECK-NEXT:    br label [[COND_END8:%.*]]
// CHECK:       cond.false7:
// CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END8]]
// CHECK:       cond.end8:
// CHECK-NEXT:    [[COND9:%.*]] = phi i32 [ 999, [[COND_TRUE6]] ], [ [[TMP25]], [[COND_FALSE7]] ]
// CHECK-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP26]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]])
// CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MIN_T1_ASCAST]], ptr [[TMP27]], align 8
// CHECK-NEXT:    %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
// CHECK-NEXT:    [[TMP28:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 4, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.13, ptr @_omp_reduction_inter_warp_copy_func.14, ptr @_omp_reduction_list_to_global_copy_func.15, ptr @_omp_reduction_list_to_global_reduce_func.16, ptr @_omp_reduction_global_to_list_copy_func.17, ptr @_omp_reduction_global_to_list_reduce_func.18)
// CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP28]], 1
// CHECK-NEXT:    br i1 [[TMP29]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP30:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-NEXT:    [[TMP31:%.*]] = load i32, ptr [[MIN_T1_ASCAST]], align 4
// CHECK-NEXT:    [[CMP10:%.*]] = icmp ult i32 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK:       cond.true11:
// CHECK-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-NEXT:    br label [[COND_END13:%.*]]
// CHECK:       cond.false12:
// CHECK-NEXT:    [[TMP33:%.*]] = load i32, ptr [[MIN_T1_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END13]]
// CHECK:       cond.end13:
// CHECK-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP32]], [[COND_TRUE11]] ], [ [[TMP33]], [[COND_FALSE12]] ]
// CHECK-NEXT:    store i32 [[COND14]], ptr [[TMP0]], align 4
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_omp_outlined_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T2:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[MIN_T2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T2]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    store i32 -1, ptr [[MIN_T2_ASCAST]], align 4
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
// CHECK-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[MIN_T2_ASCAST]], align 4
// CHECK-NEXT:    [[CONV4:%.*]] = uitofp i32 [[TMP9]] to double
// CHECK-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK-NEXT:    [[CONV5:%.*]] = uitofp i32 [[TMP12]] to double
// CHECK-NEXT:    [[TMP13:%.*]] = call nsz double @llvm.minnum.f64(double [[CONV4]], double [[CONV5]])
// CHECK-NEXT:    [[CONV6:%.*]] = fptoui double [[TMP13]] to i32
// CHECK-NEXT:    store i32 [[CONV6]], ptr [[MIN_T2_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK:       omp.body.continue:
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]])
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MIN_T2_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 4, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.11, ptr @_omp_reduction_inter_warp_copy_func.12)
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[TMP17]], 1
// CHECK-NEXT:    br i1 [[TMP18]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[MIN_T2_ASCAST]], align 4
// CHECK-NEXT:    [[CMP8:%.*]] = icmp ult i32 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[MIN_T2_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP21]], [[COND_TRUE]] ], [ [[TMP22]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i32 [[COND]], ptr [[TMP0]], align 4
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.11
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP9]], align 4
// CHECK-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
// CHECK-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
// CHECK-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP23:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
// CHECK-NEXT:    [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
// CHECK-NEXT:    [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
// CHECK-NEXT:    br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = load i32, ptr [[TMP34]], align 4
// CHECK-NEXT:    store i32 [[TMP37]], ptr [[TMP36]], align 4
// 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.12
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP9]], ptr addrspace(3) [[TMP8]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP10]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load volatile i32, ptr addrspace(3) [[TMP11]], align 4
// CHECK-NEXT:    store i32 [[TMP14]], ptr [[TMP13]], align 4
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.13
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP9]], align 4
// CHECK-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
// CHECK-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
// CHECK-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP23:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
// CHECK-NEXT:    [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
// CHECK-NEXT:    [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
// CHECK-NEXT:    br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = load i32, ptr [[TMP34]], align 4
// CHECK-NEXT:    store i32 [[TMP37]], ptr [[TMP36]], align 4
// 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.14
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP9]], ptr addrspace(3) [[TMP8]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP10]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load volatile i32, ptr addrspace(3) [[TMP11]], align 4
// CHECK-NEXT:    store i32 [[TMP14]], ptr [[TMP13]], align 4
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.15
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[TMP9]], align 4
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.16
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.17
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[TMP7]], align 4
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.18
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21
// CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR4]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_kernel_environment to ptr), ptr [[DYN_PTR]])
// CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
// CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK:       user_code.entry:
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP2]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
// CHECK-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_omp_outlined(ptr [[TMP4]], ptr [[DOTZERO_ADDR_ASCAST]], ptr [[TMP0]], ptr [[TMP3]]) #[[ATTR1]]
// CHECK-NEXT:    call void @__kmpc_target_deinit()
// CHECK-NEXT:    ret void
// CHECK:       worker.exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T1]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    store i32 0, ptr [[MAX_T1_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 999
// CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 1000
// CHECK-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP8]] to ptr
// CHECK-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
// CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP10]] to ptr
// CHECK-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
// CHECK-NEXT:    store ptr [[MAX_T1_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
// CHECK-NEXT:    store ptr [[TMP11]], ptr [[TMP17]], align 8
// CHECK-NEXT:    call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP2]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 4, i32 0)
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP24]], 999
// CHECK-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK:       cond.true6:
// CHECK-NEXT:    br label [[COND_END8:%.*]]
// CHECK:       cond.false7:
// CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END8]]
// CHECK:       cond.end8:
// CHECK-NEXT:    [[COND9:%.*]] = phi i32 [ 999, [[COND_TRUE6]] ], [ [[TMP25]], [[COND_FALSE7]] ]
// CHECK-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP26]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]])
// CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MAX_T1_ASCAST]], ptr [[TMP27]], align 8
// CHECK-NEXT:    %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
// CHECK-NEXT:    [[TMP28:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 4, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.21, ptr @_omp_reduction_inter_warp_copy_func.22, ptr @_omp_reduction_list_to_global_copy_func.23, ptr @_omp_reduction_list_to_global_reduce_func.24, ptr @_omp_reduction_global_to_list_copy_func.25, ptr @_omp_reduction_global_to_list_reduce_func.26)
// CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP28]], 1
// CHECK-NEXT:    br i1 [[TMP29]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP30:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-NEXT:    [[TMP31:%.*]] = load i32, ptr [[MAX_T1_ASCAST]], align 4
// CHECK-NEXT:    [[CMP10:%.*]] = icmp ugt i32 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK:       cond.true11:
// CHECK-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-NEXT:    br label [[COND_END13:%.*]]
// CHECK:       cond.false12:
// CHECK-NEXT:    [[TMP33:%.*]] = load i32, ptr [[MAX_T1_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END13]]
// CHECK:       cond.end13:
// CHECK-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP32]], [[COND_TRUE11]] ], [ [[TMP33]], [[COND_FALSE12]] ]
// CHECK-NEXT:    store i32 [[COND14]], ptr [[TMP0]], align 4
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_omp_outlined_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T2:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[MAX_T2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T2]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[MAX_T2_ASCAST]], align 4
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
// CHECK-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[MAX_T2_ASCAST]], align 4
// CHECK-NEXT:    [[CONV4:%.*]] = uitofp i32 [[TMP9]] to double
// CHECK-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK-NEXT:    [[CONV5:%.*]] = uitofp i32 [[TMP12]] to double
// CHECK-NEXT:    [[TMP13:%.*]] = call nsz double @llvm.maxnum.f64(double [[CONV4]], double [[CONV5]])
// CHECK-NEXT:    [[CONV6:%.*]] = fptoui double [[TMP13]] to i32
// CHECK-NEXT:    store i32 [[CONV6]], ptr [[MAX_T2_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK:       omp.body.continue:
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]])
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MAX_T2_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 4, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.19, ptr @_omp_reduction_inter_warp_copy_func.20)
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[TMP17]], 1
// CHECK-NEXT:    br i1 [[TMP18]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[MAX_T2_ASCAST]], align 4
// CHECK-NEXT:    [[CMP8:%.*]] = icmp ugt i32 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[MAX_T2_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP21]], [[COND_TRUE]] ], [ [[TMP22]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i32 [[COND]], ptr [[TMP0]], align 4
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.19
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP9]], align 4
// CHECK-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
// CHECK-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
// CHECK-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP23:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
// CHECK-NEXT:    [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
// CHECK-NEXT:    [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
// CHECK-NEXT:    br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = load i32, ptr [[TMP34]], align 4
// CHECK-NEXT:    store i32 [[TMP37]], ptr [[TMP36]], align 4
// 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.20
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP9]], ptr addrspace(3) [[TMP8]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP10]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load volatile i32, ptr addrspace(3) [[TMP11]], align 4
// CHECK-NEXT:    store i32 [[TMP14]], ptr [[TMP13]], align 4
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.21
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP9]], align 4
// CHECK-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
// CHECK-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_shuffle_int32(i32 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
// CHECK-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP23:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
// CHECK-NEXT:    [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
// CHECK-NEXT:    [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
// CHECK-NEXT:    br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = load i32, ptr [[TMP34]], align 4
// CHECK-NEXT:    store i32 [[TMP37]], ptr [[TMP36]], align 4
// 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.22
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP9]], ptr addrspace(3) [[TMP8]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP10]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = load volatile i32, ptr addrspace(3) [[TMP11]], align 4
// CHECK-NEXT:    store i32 [[TMP14]], ptr [[TMP13]], align 4
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.23
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[TMP9]], align 4
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.24
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.25
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[TMP7]], align 4
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.26
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIjEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIxEvv_l17
// CHECK-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28:![0-9]+]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    store i64 9223372036854775807, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load i64, ptr [[ARRAYIDX]], align 8
// CHECK-NEXT:    [[CONV:%.*]] = sitofp i64 [[TMP18]] to double
// CHECK-NEXT:    [[TMP19:%.*]] = load i64, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = fptosi double [[CONV]] to i64
// CHECK-NEXT:    [[XTEAM_MIN:%.*]] = call i64 @llvm.smin.i64(i64 [[TMP19]], i64 [[TMP20]])
// CHECK-NEXT:    store i64 [[XTEAM_MIN]], ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP22:%.*]] = mul i32 [[TMP21]], 1
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP27:%.*]] = load i64, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    call void @__kmpc_xteamr_l_16x64(i64 [[TMP27]], ptr [[TMP2]], ptr [[TMP25]], ptr [[TMP26]], ptr @__kmpc_rfun_min_l, ptr @__kmpc_rfun_min_lds_l, i64 9223372036854775807, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIxEvv_l21
// CHECK-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    store i64 -9223372036854775808, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load i64, ptr [[ARRAYIDX]], align 8
// CHECK-NEXT:    [[CONV:%.*]] = sitofp i64 [[TMP18]] to double
// CHECK-NEXT:    [[TMP19:%.*]] = load i64, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = fptosi double [[CONV]] to i64
// CHECK-NEXT:    [[XTEAM_MAX:%.*]] = call i64 @llvm.smax.i64(i64 [[TMP19]], i64 [[TMP20]])
// CHECK-NEXT:    store i64 [[XTEAM_MAX]], ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP22:%.*]] = mul i32 [[TMP21]], 1
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP27:%.*]] = load i64, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    call void @__kmpc_xteamr_l_16x64(i64 [[TMP27]], ptr [[TMP2]], ptr [[TMP25]], ptr [[TMP26]], ptr @__kmpc_rfun_max_l, ptr @__kmpc_rfun_max_lds_l, i64 -9223372036854775808, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17
// CHECK-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR4]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_kernel_environment to ptr), ptr [[DYN_PTR]])
// CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
// CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK:       user_code.entry:
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP2]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
// CHECK-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_omp_outlined(ptr [[TMP4]], ptr [[DOTZERO_ADDR_ASCAST]], ptr [[TMP0]], ptr [[TMP3]]) #[[ATTR1]]
// CHECK-NEXT:    call void @__kmpc_target_deinit()
// CHECK-NEXT:    ret void
// CHECK:       worker.exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T1:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T1]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    store i64 -1, ptr [[MIN_T1_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 999
// CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 1000
// CHECK-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP8]] to ptr
// CHECK-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
// CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP10]] to ptr
// CHECK-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
// CHECK-NEXT:    store ptr [[MIN_T1_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
// CHECK-NEXT:    store ptr [[TMP11]], ptr [[TMP17]], align 8
// CHECK-NEXT:    call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP2]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 4, i32 0)
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP24]], 999
// CHECK-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK:       cond.true6:
// CHECK-NEXT:    br label [[COND_END8:%.*]]
// CHECK:       cond.false7:
// CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END8]]
// CHECK:       cond.end8:
// CHECK-NEXT:    [[COND9:%.*]] = phi i32 [ 999, [[COND_TRUE6]] ], [ [[TMP25]], [[COND_FALSE7]] ]
// CHECK-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP26]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]])
// CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MIN_T1_ASCAST]], ptr [[TMP27]], align 8
// CHECK-NEXT:    %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
// CHECK-NEXT:    [[TMP28:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.29, ptr @_omp_reduction_inter_warp_copy_func.30, ptr @_omp_reduction_list_to_global_copy_func.31, ptr @_omp_reduction_list_to_global_reduce_func.32, ptr @_omp_reduction_global_to_list_copy_func.33, ptr @_omp_reduction_global_to_list_reduce_func.34)
// CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP28]], 1
// CHECK-NEXT:    br i1 [[TMP29]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP30:%.*]] = load i64, ptr [[TMP0]], align 8
// CHECK-NEXT:    [[TMP31:%.*]] = load i64, ptr [[MIN_T1_ASCAST]], align 8
// CHECK-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK:       cond.true11:
// CHECK-NEXT:    [[TMP32:%.*]] = load i64, ptr [[TMP0]], align 8
// CHECK-NEXT:    br label [[COND_END13:%.*]]
// CHECK:       cond.false12:
// CHECK-NEXT:    [[TMP33:%.*]] = load i64, ptr [[MIN_T1_ASCAST]], align 8
// CHECK-NEXT:    br label [[COND_END13]]
// CHECK:       cond.end13:
// CHECK-NEXT:    [[COND14:%.*]] = phi i64 [ [[TMP32]], [[COND_TRUE11]] ], [ [[TMP33]], [[COND_FALSE12]] ]
// CHECK-NEXT:    store i64 [[COND14]], ptr [[TMP0]], align 8
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_omp_outlined_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T2:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[MIN_T2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T2]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    store i64 -1, ptr [[MIN_T2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
// CHECK-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP9:%.*]] = load i64, ptr [[MIN_T2_ASCAST]], align 8
// CHECK-NEXT:    [[CONV4:%.*]] = uitofp i64 [[TMP9]] to double
// CHECK-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[ARRAYIDX]], align 8
// CHECK-NEXT:    [[CONV5:%.*]] = uitofp i64 [[TMP12]] to double
// CHECK-NEXT:    [[TMP13:%.*]] = call nsz double @llvm.minnum.f64(double [[CONV4]], double [[CONV5]])
// CHECK-NEXT:    [[CONV6:%.*]] = fptoui double [[TMP13]] to i64
// CHECK-NEXT:    store i64 [[CONV6]], ptr [[MIN_T2_ASCAST]], align 8
// CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK:       omp.body.continue:
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]])
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MIN_T2_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.27, ptr @_omp_reduction_inter_warp_copy_func.28)
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[TMP17]], 1
// CHECK-NEXT:    br i1 [[TMP18]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP19:%.*]] = load i64, ptr [[TMP0]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = load i64, ptr [[MIN_T2_ASCAST]], align 8
// CHECK-NEXT:    [[CMP8:%.*]] = icmp ult i64 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    [[TMP21:%.*]] = load i64, ptr [[TMP0]], align 8
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP22:%.*]] = load i64, ptr [[MIN_T2_ASCAST]], align 8
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP21]], [[COND_TRUE]] ], [ [[TMP22]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i64 [[COND]], ptr [[TMP0]], align 8
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.27
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
// CHECK-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
// CHECK-NEXT:    [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
// CHECK-NEXT:    store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP23:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
// CHECK-NEXT:    [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
// CHECK-NEXT:    [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
// CHECK-NEXT:    br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = load i64, ptr [[TMP34]], align 8
// CHECK-NEXT:    store i64 [[TMP37]], ptr [[TMP36]], 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.28
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    br label [[PRECOND:%.*]]
// CHECK:       precond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
// CHECK-NEXT:    br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
// CHECK:       body:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
// CHECK-NEXT:    [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
// CHECK-NEXT:    store i32 [[TMP18]], ptr [[TMP17]], align 4
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK-NEXT:    store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    br label [[PRECOND]]
// CHECK:       exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.29
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
// CHECK-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
// CHECK-NEXT:    [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
// CHECK-NEXT:    store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP23:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
// CHECK-NEXT:    [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
// CHECK-NEXT:    [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
// CHECK-NEXT:    br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = load i64, ptr [[TMP34]], align 8
// CHECK-NEXT:    store i64 [[TMP37]], ptr [[TMP36]], 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.30
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    br label [[PRECOND:%.*]]
// CHECK:       precond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
// CHECK-NEXT:    br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
// CHECK:       body:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
// CHECK-NEXT:    [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
// CHECK-NEXT:    store i32 [[TMP18]], ptr [[TMP17]], align 4
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK-NEXT:    store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    br label [[PRECOND]]
// CHECK:       exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.31
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i64, ptr [[TMP7]], align 8
// CHECK-NEXT:    store i64 [[TMP10]], ptr [[TMP9]], align 8
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.32
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.33
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i64, ptr [[TMP9]], align 8
// CHECK-NEXT:    store i64 [[TMP10]], ptr [[TMP7]], align 8
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.34
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l17_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21
// CHECK-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR4]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_kernel_environment to ptr), ptr [[DYN_PTR]])
// CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
// CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK:       user_code.entry:
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP2]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
// CHECK-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_omp_outlined(ptr [[TMP4]], ptr [[DOTZERO_ADDR_ASCAST]], ptr [[TMP0]], ptr [[TMP3]]) #[[ATTR1]]
// CHECK-NEXT:    call void @__kmpc_target_deinit()
// CHECK-NEXT:    ret void
// CHECK:       worker.exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T1:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T1]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    store i64 0, ptr [[MAX_T1_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 999
// CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 1000
// CHECK-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP8]] to ptr
// CHECK-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
// CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP10]] to ptr
// CHECK-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
// CHECK-NEXT:    store ptr [[MAX_T1_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
// CHECK-NEXT:    store ptr [[TMP11]], ptr [[TMP17]], align 8
// CHECK-NEXT:    call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP2]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 4, i32 0)
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP24]], 999
// CHECK-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK:       cond.true6:
// CHECK-NEXT:    br label [[COND_END8:%.*]]
// CHECK:       cond.false7:
// CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    br label [[COND_END8]]
// CHECK:       cond.end8:
// CHECK-NEXT:    [[COND9:%.*]] = phi i32 [ 999, [[COND_TRUE6]] ], [ [[TMP25]], [[COND_FALSE7]] ]
// CHECK-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP26]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]])
// CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MAX_T1_ASCAST]], ptr [[TMP27]], align 8
// CHECK-NEXT:    %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
// CHECK-NEXT:    [[TMP28:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.37, ptr @_omp_reduction_inter_warp_copy_func.38, ptr @_omp_reduction_list_to_global_copy_func.39, ptr @_omp_reduction_list_to_global_reduce_func.40, ptr @_omp_reduction_global_to_list_copy_func.41, ptr @_omp_reduction_global_to_list_reduce_func.42)
// CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP28]], 1
// CHECK-NEXT:    br i1 [[TMP29]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP30:%.*]] = load i64, ptr [[TMP0]], align 8
// CHECK-NEXT:    [[TMP31:%.*]] = load i64, ptr [[MAX_T1_ASCAST]], align 8
// CHECK-NEXT:    [[CMP10:%.*]] = icmp ugt i64 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK:       cond.true11:
// CHECK-NEXT:    [[TMP32:%.*]] = load i64, ptr [[TMP0]], align 8
// CHECK-NEXT:    br label [[COND_END13:%.*]]
// CHECK:       cond.false12:
// CHECK-NEXT:    [[TMP33:%.*]] = load i64, ptr [[MAX_T1_ASCAST]], align 8
// CHECK-NEXT:    br label [[COND_END13]]
// CHECK:       cond.end13:
// CHECK-NEXT:    [[COND14:%.*]] = phi i64 [ [[TMP32]], [[COND_TRUE11]] ], [ [[TMP33]], [[COND_FALSE12]] ]
// CHECK-NEXT:    store i64 [[COND14]], ptr [[TMP0]], align 8
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_omp_outlined_omp_outlined
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]]) #[[ATTR5]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T2:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// CHECK-NEXT:    [[MAX_T2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T2]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
// CHECK-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// CHECK-NEXT:    store i64 0, ptr [[MAX_T2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK:       omp.inner.for.cond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
// CHECK-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK:       omp.inner.for.body:
// CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP9:%.*]] = load i64, ptr [[MAX_T2_ASCAST]], align 8
// CHECK-NEXT:    [[CONV4:%.*]] = uitofp i64 [[TMP9]] to double
// CHECK-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[ARRAYIDX]], align 8
// CHECK-NEXT:    [[CONV5:%.*]] = uitofp i64 [[TMP12]] to double
// CHECK-NEXT:    [[TMP13:%.*]] = call nsz double @llvm.maxnum.f64(double [[CONV4]], double [[CONV5]])
// CHECK-NEXT:    [[CONV6:%.*]] = fptoui double [[TMP13]] to i64
// CHECK-NEXT:    store i64 [[CONV6]], ptr [[MAX_T2_ASCAST]], align 8
// CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK:       omp.body.continue:
// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK:       omp.inner.for.inc:
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// CHECK-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK:       omp.inner.for.end:
// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK:       omp.loop.exit:
// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP4]])
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    store ptr [[MAX_T2_ASCAST]], ptr [[TMP16]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.35, ptr @_omp_reduction_inter_warp_copy_func.36)
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[TMP17]], 1
// CHECK-NEXT:    br i1 [[TMP18]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
// CHECK:       .omp.reduction.then:
// CHECK-NEXT:    [[TMP19:%.*]] = load i64, ptr [[TMP0]], align 8
// CHECK-NEXT:    [[TMP20:%.*]] = load i64, ptr [[MAX_T2_ASCAST]], align 8
// CHECK-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK:       cond.true:
// CHECK-NEXT:    [[TMP21:%.*]] = load i64, ptr [[TMP0]], align 8
// CHECK-NEXT:    br label [[COND_END:%.*]]
// CHECK:       cond.false:
// CHECK-NEXT:    [[TMP22:%.*]] = load i64, ptr [[MAX_T2_ASCAST]], align 8
// CHECK-NEXT:    br label [[COND_END]]
// CHECK:       cond.end:
// CHECK-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP21]], [[COND_TRUE]] ], [ [[TMP22]], [[COND_FALSE]] ]
// CHECK-NEXT:    store i64 [[COND]], ptr [[TMP0]], align 8
// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
// CHECK:       .omp.reduction.done:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.35
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
// CHECK-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
// CHECK-NEXT:    [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
// CHECK-NEXT:    store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP23:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
// CHECK-NEXT:    [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
// CHECK-NEXT:    [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
// CHECK-NEXT:    br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = load i64, ptr [[TMP34]], align 8
// CHECK-NEXT:    store i64 [[TMP37]], ptr [[TMP36]], 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.36
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    br label [[PRECOND:%.*]]
// CHECK:       precond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
// CHECK-NEXT:    br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
// CHECK:       body:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
// CHECK-NEXT:    [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
// CHECK-NEXT:    store i32 [[TMP18]], ptr [[TMP17]], align 4
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK-NEXT:    store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    br label [[PRECOND]]
// CHECK:       exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.37
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca i64, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
// CHECK-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
// CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
// CHECK-NEXT:    [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
// CHECK-NEXT:    store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
// CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
// CHECK-NEXT:    [[TMP23:%.*]] = and i16 [[TMP5]], 1
// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
// CHECK-NEXT:    [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
// CHECK-NEXT:    [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
// CHECK-NEXT:    [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
// CHECK-NEXT:    [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
// CHECK-NEXT:    [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
// CHECK-NEXT:    br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
// CHECK:       then:
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
// CHECK-NEXT:    [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
// CHECK-NEXT:    br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
// CHECK:       then4:
// CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
// CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
// CHECK-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
// CHECK-NEXT:    [[TMP37:%.*]] = load i64, ptr [[TMP34]], align 8
// CHECK-NEXT:    store i64 [[TMP37]], ptr [[TMP36]], 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.38
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
// CHECK-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    br label [[PRECOND:%.*]]
// CHECK:       precond:
// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
// CHECK-NEXT:    br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
// CHECK:       body:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
// 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:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
// CHECK-NEXT:    store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
// CHECK-NEXT:    br label [[IFCONT:%.*]]
// CHECK:       else:
// CHECK-NEXT:    br label [[IFCONT]]
// CHECK:       ifcont:
// CHECK-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
// CHECK:       then3:
// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
// CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
// CHECK-NEXT:    [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
// CHECK-NEXT:    store i32 [[TMP18]], ptr [[TMP17]], align 4
// CHECK-NEXT:    br label [[IFCONT5:%.*]]
// CHECK:       else4:
// CHECK-NEXT:    br label [[IFCONT5]]
// CHECK:       ifcont5:
// CHECK-NEXT:    [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK-NEXT:    store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
// CHECK-NEXT:    br label [[PRECOND]]
// CHECK:       exit:
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.39
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_13:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_13]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i64, ptr [[TMP7]], align 8
// CHECK-NEXT:    store i64 [[TMP10]], ptr [[TMP9]], align 8
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.40
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_13:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_13]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.41
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_13:%.*]], ptr [[TMP4]], i32 [[TMP5]]
// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_13]], ptr [[TMP8]], i32 0, i32 0
// CHECK-NEXT:    [[TMP10:%.*]] = load i64, ptr [[TMP9]], align 8
// CHECK-NEXT:    store i64 [[TMP10]], ptr [[TMP7]], align 8
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.42
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR6]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_13:%.*]], ptr [[TMP3]], i32 [[TMP4]]
// CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_13]], ptr [[TMP6]], i32 0, i32 0
// CHECK-NEXT:    store ptr [[TMP7]], ptr [[TMP5]], align 8
// CHECK-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
// CHECK-NEXT:    call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIyEvv_l21_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR1]]
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIfEvv_l17
// CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT:    store float +inf, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load float, ptr [[ARRAYIDX]], align 4
// CHECK-NEXT:    [[CONV:%.*]] = fpext float [[TMP18]] to double
// CHECK-NEXT:    [[TMP19:%.*]] = load float, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = fptrunc double [[CONV]] to float
// CHECK-NEXT:    [[XTEAM_MIN:%.*]] = call float @llvm.minnum.f32(float [[TMP19]], float [[TMP20]])
// CHECK-NEXT:    store float [[XTEAM_MIN]], ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP22:%.*]] = mul i32 [[TMP21]], 1
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP27:%.*]] = load float, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_xteamr_f_16x64(float [[TMP27]], ptr [[TMP2]], ptr [[TMP25]], ptr [[TMP26]], ptr @__kmpc_rfun_min_f, ptr @__kmpc_rfun_min_lds_f, float +inf, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIfEvv_l21
// CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META25]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT:    store float -inf, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load float, ptr [[ARRAYIDX]], align 4
// CHECK-NEXT:    [[CONV:%.*]] = fpext float [[TMP18]] to double
// CHECK-NEXT:    [[TMP19:%.*]] = load float, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    [[TMP20:%.*]] = fptrunc double [[CONV]] to float
// CHECK-NEXT:    [[XTEAM_MAX:%.*]] = call float @llvm.maxnum.f32(float [[TMP19]], float [[TMP20]])
// CHECK-NEXT:    store float [[XTEAM_MAX]], ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP22:%.*]] = mul i32 [[TMP21]], 1
// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
// CHECK-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP27:%.*]] = load float, ptr addrspace(5) [[TMP3]], align 4
// CHECK-NEXT:    call void @__kmpc_xteamr_f_16x64(float [[TMP27]], ptr [[TMP2]], ptr [[TMP25]], ptr [[TMP26]], ptr @__kmpc_rfun_max_f, ptr @__kmpc_rfun_max_lds_f, float -inf, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIdEvv_l17
// CHECK-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[MIN_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MIN_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MIN_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MIN_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MIN_T]], ptr [[MIN_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MIN_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT:    store double +inf, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load double, ptr [[ARRAYIDX]], align 8
// CHECK-NEXT:    [[TMP19:%.*]] = load double, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    [[XTEAM_MIN:%.*]] = call double @llvm.minnum.f64(double [[TMP19]], double [[TMP18]])
// CHECK-NEXT:    store double [[XTEAM_MIN]], ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP20:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[TMP20]], 1
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP23:%.*]] = add i32 [[TMP21]], [[TMP22]]
// CHECK-NEXT:    store i32 [[TMP23]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load double, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    call void @__kmpc_xteamr_d_16x64(double [[TMP26]], ptr [[TMP2]], ptr [[TMP24]], ptr [[TMP25]], ptr @__kmpc_rfun_min_d, ptr @__kmpc_rfun_min_lds_d, double +inf, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15compute_min_maxIdEvv_l21
// CHECK-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[MAX_T:%.*]], ptr noundef [[ARR_T:%.*]], ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[MAX_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[ARR_T_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT:    [[MAX_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[MAX_T_ADDR]] to ptr
// CHECK-NEXT:    [[ARR_T_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ARR_T_ADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
// CHECK-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// CHECK-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// CHECK-NEXT:    store ptr [[MAX_T]], ptr [[MAX_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[ARR_T]], ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MAX_T_ADDR_ASCAST]], align 8, !nonnull [[META20]], !align [[META28]]
// CHECK-NEXT:    call void @__kmpc_specialized_kernel_init()
// CHECK-NEXT:    [[TMP3:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT:    store double -inf, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 999, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[GPU_BLOCK_ID:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
// CHECK-NEXT:    [[TMP6:%.*]] = mul i32 [[GPU_BLOCK_ID]], [[NVPTX_NUM_THREADS]]
// CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
// CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 1
// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[TMP8]], [[TMP9]]
// CHECK-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_get_hardware_num_blocks()
// CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP10]] to i64
// CHECK-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND:%.*]]
// CHECK:       for.cond:
// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK:       for.body:
// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-NEXT:    store i32 [[ADD]], ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[ARR_T_ADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[TMP16]], i64 [[IDXPROM]]
// CHECK-NEXT:    [[TMP18:%.*]] = load double, ptr [[ARRAYIDX]], align 8
// CHECK-NEXT:    [[TMP19:%.*]] = load double, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    [[XTEAM_MAX:%.*]] = call double @llvm.maxnum.f64(double [[TMP19]], double [[TMP18]])
// CHECK-NEXT:    store double [[XTEAM_MAX]], ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    br label [[FOR_INC:%.*]]
// CHECK:       for.inc:
// CHECK-NEXT:    [[NVPTX_NUM_THREADS2:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-NEXT:    [[TMP20:%.*]] = mul i32 [[NVPTX_NUM_THREADS2]], [[TMP11]]
// CHECK-NEXT:    [[TMP21:%.*]] = mul i32 [[TMP20]], 1
// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    [[TMP23:%.*]] = add i32 [[TMP21]], [[TMP22]]
// CHECK-NEXT:    store i32 [[TMP23]], ptr [[DOTOMP_IV_ASCAST]], align 4
// CHECK-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK:       for.end:
// CHECK-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
// CHECK-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTADDR1_ASCAST]], align 8
// CHECK-NEXT:    [[TMP26:%.*]] = load double, ptr addrspace(5) [[TMP3]], align 8
// CHECK-NEXT:    call void @__kmpc_xteamr_d_16x64(double [[TMP26]], ptr [[TMP2]], ptr [[TMP24]], ptr [[TMP25]], ptr @__kmpc_rfun_max_d, ptr @__kmpc_rfun_max_lds_d, double -inf, i64 [[TMP12]], i32 [[TMP11]], i32 1)
// CHECK-NEXT:    ret void
//