// 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 _
// Test target codegen - host bc file has to be created first.
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple nvptx64-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck --check-prefixes=CHECK-64 %s
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm-bc %s -o %t-x86-host.bc
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck --check-prefixes=CHECK-32 %s
// RUN: %clang_cc1 -verify -fopenmp -fexceptions -fcxx-exceptions -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck --check-prefixes=CHECK-32-EX %s
// expected-no-diagnostics
#ifndef HEADER
#define HEADER

int a;

void foo() {
#pragma omp target teams distribute parallel for simd if(a)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for simd schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for simd schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for simd schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for simd schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for simd schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for simd schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
int a;
#pragma omp target teams distribute parallel for lastprivate(a)
  for (int i = 0; i < 10; ++i)
    a = i;
#pragma omp target teams distribute parallel for schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams distribute parallel for schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
   {
     int b;
#pragma omp distribute parallel for simd
  for (int i = 0; i < 10; ++i)
    ;
  ;
   }
#pragma omp target teams
   {
     int b[] = {2, 3, sizeof(int)};
#pragma omp distribute parallel for simd schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
   }
#pragma omp target teams
   {
     int b;
#pragma omp distribute parallel for simd schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
  int &c = b;
   }
#pragma omp target teams
#pragma omp distribute parallel for simd schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for simd schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for simd schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for simd schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target teams
#pragma omp distribute parallel for schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel for if(a)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel for schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel for schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel for schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel for schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel for schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel for schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel if(a)
#pragma omp for simd
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel
#pragma omp for simd schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel
#pragma omp for simd schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel
#pragma omp for simd schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel
#pragma omp for simd schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel
#pragma omp for simd schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target parallel
#pragma omp for simd schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel
#pragma omp for simd ordered
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel
#pragma omp for simd schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel
#pragma omp for simd schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel
#pragma omp for simd schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel
#pragma omp for simd schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel
#pragma omp for simd schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel
#pragma omp for simd schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel for
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel for schedule(static)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel for schedule(static, 1)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel for schedule(auto)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel for schedule(runtime)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel for schedule(dynamic)
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp target
#pragma omp parallel for schedule(guided)
  for (int i = 0; i < 10; ++i)
    ;
}

#endif
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-64-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
// CHECK-64-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i64 [[TMP3]]) #[[ATTR2:[0-9]+]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x ptr], align 8
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__CASTED15:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS17:%.*]] = alloca [3 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP5]] to i1
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-64:       omp_if.then:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP11]] to i1
// CHECK-64-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL2]] to i8
// CHECK-64-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP14:%.*]] = inttoptr i64 [[TMP8]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP14]], ptr [[TMP13]], align 8, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP16:%.*]] = inttoptr i64 [[TMP10]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP16]], ptr [[TMP15]], align 8, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
// CHECK-64-NEXT:    [[TMP18:%.*]] = inttoptr i64 [[TMP12]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP18]], ptr [[TMP17]], align 8, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP19]] to i1
// CHECK-64-NEXT:    [[TMP20:%.*]] = zext i1 [[TOBOOL3]] to i32
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP20]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 3), !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK-64-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP27]], 9
// CHECK-64-NEXT:    br i1 [[CMP6]], label [[COND_TRUE7:%.*]], label [[COND_FALSE8:%.*]]
// CHECK-64:       cond.true7:
// CHECK-64-NEXT:    br label [[COND_END9:%.*]]
// CHECK-64:       cond.false8:
// CHECK-64-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    br label [[COND_END9]]
// CHECK-64:       cond.end9:
// CHECK-64-NEXT:    [[COND10:%.*]] = phi i32 [ 9, [[COND_TRUE7]] ], [ [[TMP28]], [[COND_FALSE8]] ]
// CHECK-64-NEXT:    store i32 [[COND10]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    store i32 [[TMP29]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP222:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-64:       omp_if.else:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND11:%.*]]
// CHECK-64:       omp.inner.for.cond11:
// CHECK-64-NEXT:    [[TMP30:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP12:%.*]] = icmp slt i32 [[TMP30]], 10
// CHECK-64-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY13:%.*]], label [[OMP_INNER_FOR_END28:%.*]]
// CHECK-64:       omp.inner.for.body13:
// CHECK-64-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP32:%.*]] = zext i32 [[TMP31]] to i64
// CHECK-64-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP34:%.*]] = zext i32 [[TMP33]] to i64
// CHECK-64-NEXT:    [[TMP35:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-64-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP35]] to i1
// CHECK-64-NEXT:    [[FROMBOOL16:%.*]] = zext i1 [[TOBOOL14]] to i8
// CHECK-64-NEXT:    store i8 [[FROMBOOL16]], ptr [[DOTCAPTURE_EXPR__CASTED15]], align 1
// CHECK-64-NEXT:    [[TMP36:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED15]], align 8
// CHECK-64-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP38:%.*]] = inttoptr i64 [[TMP32]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP38]], ptr [[TMP37]], align 8
// CHECK-64-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP40:%.*]] = inttoptr i64 [[TMP34]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP40]], ptr [[TMP39]], align 8
// CHECK-64-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i64 0, i64 2
// CHECK-64-NEXT:    [[TMP42:%.*]] = inttoptr i64 [[TMP36]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP42]], ptr [[TMP41]], align 8
// CHECK-64-NEXT:    [[TMP43:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-64-NEXT:    [[TOBOOL18:%.*]] = trunc i8 [[TMP43]] to i1
// CHECK-64-NEXT:    [[TMP44:%.*]] = zext i1 [[TOBOOL18]] to i32
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP44]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined1, ptr null, ptr [[CAPTURED_VARS_ADDRS17]], i64 3)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC19:%.*]]
// CHECK-64:       omp.inner.for.inc19:
// CHECK-64-NEXT:    [[TMP45:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP46:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP45]], [[TMP46]]
// CHECK-64-NEXT:    store i32 [[ADD20]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP47:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP48:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP47]], [[TMP48]]
// CHECK-64-NEXT:    store i32 [[ADD21]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP49:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP50:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD22:%.*]] = add nsw i32 [[TMP49]], [[TMP50]]
// CHECK-64-NEXT:    store i32 [[ADD22]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP51:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP23:%.*]] = icmp sgt i32 [[TMP51]], 9
// CHECK-64-NEXT:    br i1 [[CMP23]], label [[COND_TRUE24:%.*]], label [[COND_FALSE25:%.*]]
// CHECK-64:       cond.true24:
// CHECK-64-NEXT:    br label [[COND_END26:%.*]]
// CHECK-64:       cond.false25:
// CHECK-64-NEXT:    [[TMP52:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END26]]
// CHECK-64:       cond.end26:
// CHECK-64-NEXT:    [[COND27:%.*]] = phi i32 [ 9, [[COND_TRUE24]] ], [ [[TMP52]], [[COND_FALSE25]] ]
// CHECK-64-NEXT:    store i32 [[COND27]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP53:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP53]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND11]], !llvm.loop [[LOOP225:![0-9]+]]
// CHECK-64:       omp.inner.for.end28:
// CHECK-64-NEXT:    br label [[OMP_IF_END]]
// CHECK-64:       omp_if.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP54:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP55:%.*]] = icmp ne i32 [[TMP54]], 0
// CHECK-64-NEXT:    br i1 [[TMP55]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-64:       omp_if.then:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227:![0-9]+]]
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP227]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP228:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-64:       omp_if.else:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP12]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND4:%.*]]
// CHECK-64:       omp.inner.for.cond4:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP14]] to i64
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP15]]
// CHECK-64-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY7:%.*]], label [[OMP_INNER_FOR_END13:%.*]]
// CHECK-64:       omp.inner.for.body7:
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK-64-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK-64-NEXT:    store i32 [[ADD9]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE10:%.*]]
// CHECK-64:       omp.body.continue10:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC11:%.*]]
// CHECK-64:       omp.inner.for.inc11:
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK-64-NEXT:    store i32 [[ADD12]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND4]], !llvm.loop [[LOOP230:![0-9]+]]
// CHECK-64:       omp.inner.for.end13:
// CHECK-64-NEXT:    br label [[OMP_IF_END]]
// CHECK-64:       omp_if.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP20]])
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-64-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined1
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-64:       omp_if.then:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231:![0-9]+]]
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP231]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP232:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-64:       omp_if.else:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP12]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND4:%.*]]
// CHECK-64:       omp.inner.for.cond4:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP14]] to i64
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP15]]
// CHECK-64-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY7:%.*]], label [[OMP_INNER_FOR_END13:%.*]]
// CHECK-64:       omp.inner.for.body7:
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK-64-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
// CHECK-64-NEXT:    store i32 [[ADD9]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE10:%.*]]
// CHECK-64:       omp.body.continue10:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC11:%.*]]
// CHECK-64:       omp.inner.for.inc11:
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK-64-NEXT:    store i32 [[ADD12]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND4]], !llvm.loop [[LOOP234:![0-9]+]]
// CHECK-64:       omp.inner.for.end13:
// CHECK-64-NEXT:    br label [[OMP_IF_END]]
// CHECK-64:       omp_if.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP20]])
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-64-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP236:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238:![0-9]+]]
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP239:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-64-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP242:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244:![0-9]+]]
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP244]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP245:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-64-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP248:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP251:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP254:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP257:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP260:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP263:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP266:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP269:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[A:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP2]], ptr [[A_CASTED]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i64, ptr [[A_CASTED]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i64 [[TMP3]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK-64-NEXT:    [[A1:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i64 4)
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[A1]], align 4
// CHECK-64-NEXT:    store i32 [[TMP10]], ptr [[A_CASTED]], align 4
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i64, ptr [[A_CASTED]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 8
// CHECK-64-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
// CHECK-64-NEXT:    [[TMP17:%.*]] = inttoptr i64 [[TMP11]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP17]], ptr [[TMP16]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 3)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP24]], 9
// CHECK-64-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK-64:       cond.true6:
// CHECK-64-NEXT:    br label [[COND_END8:%.*]]
// CHECK-64:       cond.false7:
// CHECK-64-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END8]]
// CHECK-64:       cond.end8:
// CHECK-64-NEXT:    [[COND9:%.*]] = phi i32 [ 9, [[COND_TRUE6]] ], [ [[TMP25]], [[COND_FALSE7]] ]
// CHECK-64-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP26]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
// CHECK-64-NEXT:    br i1 [[TMP28]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
// CHECK-64:       .omp.lastprivate.then:
// CHECK-64-NEXT:    [[TMP29:%.*]] = load i32, ptr [[A1]], align 4
// CHECK-64-NEXT:    store i32 [[TMP29]], ptr [[A_ADDR]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
// CHECK-64:       .omp.lastprivate.done:
// CHECK-64-NEXT:    call void @__kmpc_free_shared(ptr [[A1]], i64 4)
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[A2:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[I]], align 4
// CHECK-64-NEXT:    store i32 [[TMP8]], ptr [[A2]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-64-NEXT:    br i1 [[TMP12]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
// CHECK-64:       .omp.lastprivate.then:
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[A2]], align 4
// CHECK-64-NEXT:    store i32 [[TMP13]], ptr [[A_ADDR]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
// CHECK-64:       .omp.lastprivate.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP272:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP275:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP278:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP281:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP284:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286:![0-9]+]]
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP286]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP287:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-64-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[B:%.*]] = alloca [3 x i32], align 4
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[B]], ptr align 4 @"__const.<captured>.b", i64 12, i1 false)
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP290:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292:![0-9]+]]
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP293:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-64-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[C:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i64 8)
// CHECK-64-NEXT:    [[B:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i64 4)
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295:![0-9]+]]
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK-64-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP8]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP12]], ptr [[TMP11]], align 8, !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP14:%.*]] = inttoptr i64 [[TMP10]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP14]], ptr [[TMP13]], align 8, !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP296:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = icmp ne i32 [[TMP17]], 0
// CHECK-64-NEXT:    br i1 [[TMP18]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    store ptr [[B]], ptr [[C]], align 8
// CHECK-64-NEXT:    call void @__kmpc_free_shared(ptr [[B]], i64 4)
// CHECK-64-NEXT:    call void @__kmpc_free_shared(ptr [[C]], i64 8)
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298:![0-9]+]]
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP298]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP299:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-64-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined_wrapper
// CHECK-64-SAME: (i16 noundef zeroext [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR7:[0-9]+]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTADDR:%.*]] = alloca i16, align 2
// CHECK-64-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    store i16 [[TMP0]], ptr [[DOTADDR]], align 2
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = getelementptr inbounds ptr, ptr [[TMP2]], i64 0
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = getelementptr inbounds ptr, ptr [[TMP2]], i64 1
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[TMP5]], align 8
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], i64 [[TMP4]], i64 [[TMP6]]) #[[ATTR2]]
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP302:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP305:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP308:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP311:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP314:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP317:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319:![0-9]+]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2), !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP320:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
// CHECK-64-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP323:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-64-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP326:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP329:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP332:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP335:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP338:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP341:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP344:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-64-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-64-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP6]] to i64
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-64-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
// CHECK-64-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP7]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
// CHECK-64-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
// CHECK-64-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP9]] to ptr
// CHECK-64-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 2)
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP20]], 9
// CHECK-64-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-64:       cond.true5:
// CHECK-64-NEXT:    br label [[COND_END7:%.*]]
// CHECK-64:       cond.false6:
// CHECK-64-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END7]]
// CHECK-64:       cond.end7:
// CHECK-64-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP21]], [[COND_FALSE6]] ]
// CHECK-64-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-64-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP22]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK-64-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK-64-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346:![0-9]+]]
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP347:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR8:[0-9]+]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-64-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i32
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP3]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-64-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-64-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP350:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP353:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP356:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP359:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-64-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i32
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP3]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361:![0-9]+]]
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP362:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-64-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-64-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4:[0-9]+]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364:![0-9]+]]
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP365:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
// CHECK-64-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367:![0-9]+]]
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP368:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-64-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-64-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP371:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP374:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP377:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP380:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10:[0-9]+]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 65, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-64-NEXT:    call void @__kmpc_dispatch_fini_4(ptr @[[GLOB1]], i32 [[TMP1]]), !llvm.access.group [[ACC_GRP382]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP383:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385:![0-9]+]]
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP386:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
// CHECK-64-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388:![0-9]+]]
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP389:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-64-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-64-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP392:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP395:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP398:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP401:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-64-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-64:       .omp.final.then:
// CHECK-64-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-64:       .omp.final.done:
// CHECK-64-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-64-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-64-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-64:       omp.loop.exit:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-64-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-64:       cond.true:
// CHECK-64-NEXT:    br label [[COND_END:%.*]]
// CHECK-64:       cond.false:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[COND_END]]
// CHECK-64:       cond.end:
// CHECK-64-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-64-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-64-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-64-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-64-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-64-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-64-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP404:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP407:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP410:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268
// CHECK-64-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8
// CHECK-64-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK-64-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_kernel_environment, ptr [[DYN_PTR]])
// CHECK-64-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-64-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-64:       user_code.entry:
// CHECK-64-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-64-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 0)
// CHECK-64-NEXT:    call void @__kmpc_target_deinit()
// CHECK-64-NEXT:    ret void
// CHECK-64:       worker.exit:
// CHECK-64-NEXT:    ret void
//
//
// CHECK-64-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_omp_outlined
// CHECK-64-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-64-NEXT:  entry:
// CHECK-64-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK-64-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-64-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-64-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-64-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-64-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK-64-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-64-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-64:       omp.dispatch.cond:
// CHECK-64-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-64-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-64-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-64:       omp.dispatch.body:
// CHECK-64-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-64-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64:       omp.inner.for.cond:
// CHECK-64-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412:![0-9]+]]
// CHECK-64-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-64-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-64-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64:       omp.inner.for.body:
// CHECK-64-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-64-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-64-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-64-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-64-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-64:       omp.body.continue:
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-64:       omp.inner.for.inc:
// CHECK-64-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-64-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-64-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-64-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP413:![0-9]+]]
// CHECK-64:       omp.inner.for.end:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-64:       omp.dispatch.inc:
// CHECK-64-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-64:       omp.dispatch.end:
// CHECK-64-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-64-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
// CHECK-32-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR2:[0-9]+]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x ptr], align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__CASTED15:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS17:%.*]] = alloca [3 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP5]] to i1
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-32:       omp_if.then:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
// CHECK-32-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL2]] to i8
// CHECK-32-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP12]], ptr [[TMP11]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP14:%.*]] = inttoptr i32 [[TMP8]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP14]], ptr [[TMP13]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
// CHECK-32-NEXT:    [[TMP16:%.*]] = inttoptr i32 [[TMP10]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP16]], ptr [[TMP15]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP17]] to i1
// CHECK-32-NEXT:    [[TMP18:%.*]] = zext i1 [[TOBOOL3]] to i32
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP18]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 3), !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK-32-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP25]], 9
// CHECK-32-NEXT:    br i1 [[CMP6]], label [[COND_TRUE7:%.*]], label [[COND_FALSE8:%.*]]
// CHECK-32:       cond.true7:
// CHECK-32-NEXT:    br label [[COND_END9:%.*]]
// CHECK-32:       cond.false8:
// CHECK-32-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    br label [[COND_END9]]
// CHECK-32:       cond.end9:
// CHECK-32-NEXT:    [[COND10:%.*]] = phi i32 [ 9, [[COND_TRUE7]] ], [ [[TMP26]], [[COND_FALSE8]] ]
// CHECK-32-NEXT:    store i32 [[COND10]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    store i32 [[TMP27]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP222:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-32:       omp_if.else:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND11:%.*]]
// CHECK-32:       omp.inner.for.cond11:
// CHECK-32-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP12:%.*]] = icmp slt i32 [[TMP28]], 10
// CHECK-32-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY13:%.*]], label [[OMP_INNER_FOR_END28:%.*]]
// CHECK-32:       omp.inner.for.body13:
// CHECK-32-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP30:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP31:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP31]] to i1
// CHECK-32-NEXT:    [[FROMBOOL16:%.*]] = zext i1 [[TOBOOL14]] to i8
// CHECK-32-NEXT:    store i8 [[FROMBOOL16]], ptr [[DOTCAPTURE_EXPR__CASTED15]], align 1
// CHECK-32-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED15]], align 4
// CHECK-32-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP34:%.*]] = inttoptr i32 [[TMP29]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP34]], ptr [[TMP33]], align 4
// CHECK-32-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP36:%.*]] = inttoptr i32 [[TMP30]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP36]], ptr [[TMP35]], align 4
// CHECK-32-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i32 0, i32 2
// CHECK-32-NEXT:    [[TMP38:%.*]] = inttoptr i32 [[TMP32]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP38]], ptr [[TMP37]], align 4
// CHECK-32-NEXT:    [[TMP39:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-NEXT:    [[TOBOOL18:%.*]] = trunc i8 [[TMP39]] to i1
// CHECK-32-NEXT:    [[TMP40:%.*]] = zext i1 [[TOBOOL18]] to i32
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP40]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined1, ptr null, ptr [[CAPTURED_VARS_ADDRS17]], i32 3)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC19:%.*]]
// CHECK-32:       omp.inner.for.inc19:
// CHECK-32-NEXT:    [[TMP41:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP42:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP41]], [[TMP42]]
// CHECK-32-NEXT:    store i32 [[ADD20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP43:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP44:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP43]], [[TMP44]]
// CHECK-32-NEXT:    store i32 [[ADD21]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP45:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP46:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD22:%.*]] = add nsw i32 [[TMP45]], [[TMP46]]
// CHECK-32-NEXT:    store i32 [[ADD22]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP47:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP23:%.*]] = icmp sgt i32 [[TMP47]], 9
// CHECK-32-NEXT:    br i1 [[CMP23]], label [[COND_TRUE24:%.*]], label [[COND_FALSE25:%.*]]
// CHECK-32:       cond.true24:
// CHECK-32-NEXT:    br label [[COND_END26:%.*]]
// CHECK-32:       cond.false25:
// CHECK-32-NEXT:    [[TMP48:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END26]]
// CHECK-32:       cond.end26:
// CHECK-32-NEXT:    [[COND27:%.*]] = phi i32 [ 9, [[COND_TRUE24]] ], [ [[TMP48]], [[COND_FALSE25]] ]
// CHECK-32-NEXT:    store i32 [[COND27]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP49:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP49]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND11]], !llvm.loop [[LOOP225:![0-9]+]]
// CHECK-32:       omp.inner.for.end28:
// CHECK-32-NEXT:    br label [[OMP_IF_END]]
// CHECK-32:       omp_if.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP50:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP51:%.*]] = icmp ne i32 [[TMP50]], 0
// CHECK-32-NEXT:    br i1 [[TMP51]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-32:       omp_if.then:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227:![0-9]+]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP228:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-32:       omp_if.else:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP12]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND2:%.*]]
// CHECK-32:       omp.inner.for.cond2:
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    [[CMP3:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY4:%.*]], label [[OMP_INNER_FOR_END10:%.*]]
// CHECK-32:       omp.inner.for.body4:
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK-32-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
// CHECK-32-NEXT:    store i32 [[ADD6]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE7:%.*]]
// CHECK-32:       omp.body.continue7:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC8:%.*]]
// CHECK-32:       omp.inner.for.inc8:
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK-32-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND2]], !llvm.loop [[LOOP230:![0-9]+]]
// CHECK-32:       omp.inner.for.end10:
// CHECK-32-NEXT:    br label [[OMP_IF_END]]
// CHECK-32:       omp_if.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP20]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined1
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-32:       omp_if.then:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231:![0-9]+]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP232:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-32:       omp_if.else:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP12]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND2:%.*]]
// CHECK-32:       omp.inner.for.cond2:
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    [[CMP3:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY4:%.*]], label [[OMP_INNER_FOR_END10:%.*]]
// CHECK-32:       omp.inner.for.body4:
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK-32-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
// CHECK-32-NEXT:    store i32 [[ADD6]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE7:%.*]]
// CHECK-32:       omp.body.continue7:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC8:%.*]]
// CHECK-32:       omp.inner.for.inc8:
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK-32-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND2]], !llvm.loop [[LOOP234:![0-9]+]]
// CHECK-32:       omp.inner.for.end10:
// CHECK-32-NEXT:    br label [[OMP_IF_END]]
// CHECK-32:       omp_if.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP20]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP236:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238:![0-9]+]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP239:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-32-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP242:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244:![0-9]+]]
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP245:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-32-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP248:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP251:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP254:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP257:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP260:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP263:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP266:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP269:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[A:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP2]], ptr [[A_CASTED]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A_CASTED]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK-32-NEXT:    [[A1:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i32 4)
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A1]], align 4
// CHECK-32-NEXT:    store i32 [[TMP8]], ptr [[A_CASTED]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[A_CASTED]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP13:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
// CHECK-32-NEXT:    [[TMP15:%.*]] = inttoptr i32 [[TMP9]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 3)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP22]], 9
// CHECK-32-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK-32:       cond.true6:
// CHECK-32-NEXT:    br label [[COND_END8:%.*]]
// CHECK-32:       cond.false7:
// CHECK-32-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END8]]
// CHECK-32:       cond.end8:
// CHECK-32-NEXT:    [[COND9:%.*]] = phi i32 [ 9, [[COND_TRUE6]] ], [ [[TMP23]], [[COND_FALSE7]] ]
// CHECK-32-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK-32-NEXT:    br i1 [[TMP26]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
// CHECK-32:       .omp.lastprivate.then:
// CHECK-32-NEXT:    [[TMP27:%.*]] = load i32, ptr [[A1]], align 4
// CHECK-32-NEXT:    store i32 [[TMP27]], ptr [[A_ADDR]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
// CHECK-32:       .omp.lastprivate.done:
// CHECK-32-NEXT:    call void @__kmpc_free_shared(ptr [[A1]], i32 4)
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[A1:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[I]], align 4
// CHECK-32-NEXT:    store i32 [[TMP8]], ptr [[A1]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-32-NEXT:    br i1 [[TMP12]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
// CHECK-32:       .omp.lastprivate.then:
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[A1]], align 4
// CHECK-32-NEXT:    store i32 [[TMP13]], ptr [[A_ADDR]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
// CHECK-32:       .omp.lastprivate.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP272:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP275:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP278:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP281:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP284:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286:![0-9]+]]
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP287:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-32-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[B:%.*]] = alloca [3 x i32], align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[B]], ptr align 4 @"__const.<captured>.b", i32 12, i1 false)
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP290:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292:![0-9]+]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP293:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-32-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[C:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i32 4)
// CHECK-32-NEXT:    [[B:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i32 4)
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295:![0-9]+]]
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP10:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP10]], ptr [[TMP9]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP8]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP12]], ptr [[TMP11]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP296:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-32-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    store ptr [[B]], ptr [[C]], align 4
// CHECK-32-NEXT:    call void @__kmpc_free_shared(ptr [[B]], i32 4)
// CHECK-32-NEXT:    call void @__kmpc_free_shared(ptr [[C]], i32 4)
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298:![0-9]+]]
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP299:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-32-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined_wrapper
// CHECK-32-SAME: (i16 noundef zeroext [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR7:[0-9]+]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTADDR:%.*]] = alloca i16, align 2
// CHECK-32-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    store i16 [[TMP0]], ptr [[DOTADDR]], align 2
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = getelementptr inbounds ptr, ptr [[TMP2]], i32 0
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = getelementptr inbounds ptr, ptr [[TMP2]], i32 1
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], i32 [[TMP4]], i32 [[TMP6]]) #[[ATTR2]]
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP302:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP305:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP308:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP311:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP314:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP317:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319:![0-9]+]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP320:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP323:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP326:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP329:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP332:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP335:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP338:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP341:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP344:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32:       cond.true5:
// CHECK-32-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32:       cond.false6:
// CHECK-32-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END7]]
// CHECK-32:       cond.end7:
// CHECK-32-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346:![0-9]+]]
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP347:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR8:[0-9]+]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i32
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP3]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP350:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP353:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP356:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP359:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i32
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP3]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361:![0-9]+]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP362:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-32-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4:[0-9]+]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364:![0-9]+]]
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP365:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
// CHECK-32-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367:![0-9]+]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP368:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-32-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP371:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP374:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP377:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP380:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10:[0-9]+]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 65, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-NEXT:    call void @__kmpc_dispatch_fini_4(ptr @[[GLOB1]], i32 [[TMP1]]), !llvm.access.group [[ACC_GRP382]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP383:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385:![0-9]+]]
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP386:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
// CHECK-32-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388:![0-9]+]]
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP389:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-32-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP392:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP395:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP398:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP401:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32:       .omp.final.then:
// CHECK-32-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32:       .omp.final.done:
// CHECK-32-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-32-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32:       omp.loop.exit:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32:       cond.true:
// CHECK-32-NEXT:    br label [[COND_END:%.*]]
// CHECK-32:       cond.false:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[COND_END]]
// CHECK-32:       cond.end:
// CHECK-32-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP404:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP407:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP410:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268
// CHECK-32-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32:       user_code.entry:
// CHECK-32-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-NEXT:    ret void
// CHECK-32:       worker.exit:
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_omp_outlined
// CHECK-32-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-NEXT:  entry:
// CHECK-32-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32:       omp.dispatch.cond:
// CHECK-32-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32:       omp.dispatch.body:
// CHECK-32-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32:       omp.inner.for.cond:
// CHECK-32-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412:![0-9]+]]
// CHECK-32-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32:       omp.inner.for.body:
// CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32:       omp.body.continue:
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32:       omp.inner.for.inc:
// CHECK-32-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP413:![0-9]+]]
// CHECK-32:       omp.inner.for.end:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32:       omp.dispatch.inc:
// CHECK-32-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32:       omp.dispatch.end:
// CHECK-32-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-EX-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
// CHECK-32-EX-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR2:[0-9]+]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x ptr], align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__CASTED15:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS17:%.*]] = alloca [3 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP5]] to i1
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-32-EX:       omp_if.then:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
// CHECK-32-EX-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL2]] to i8
// CHECK-32-EX-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP12]], ptr [[TMP11]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = inttoptr i32 [[TMP8]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP14]], ptr [[TMP13]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = inttoptr i32 [[TMP10]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP16]], ptr [[TMP15]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP17]] to i1
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = zext i1 [[TOBOOL3]] to i32
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP18]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 3), !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
// CHECK-32-EX-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP25]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP6]], label [[COND_TRUE7:%.*]], label [[COND_FALSE8:%.*]]
// CHECK-32-EX:       cond.true7:
// CHECK-32-EX-NEXT:    br label [[COND_END9:%.*]]
// CHECK-32-EX:       cond.false8:
// CHECK-32-EX-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    br label [[COND_END9]]
// CHECK-32-EX:       cond.end9:
// CHECK-32-EX-NEXT:    [[COND10:%.*]] = phi i32 [ 9, [[COND_TRUE7]] ], [ [[TMP26]], [[COND_FALSE8]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND10]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    store i32 [[TMP27]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP221]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP222:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-32-EX:       omp_if.else:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND11:%.*]]
// CHECK-32-EX:       omp.inner.for.cond11:
// CHECK-32-EX-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP12:%.*]] = icmp slt i32 [[TMP28]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY13:%.*]], label [[OMP_INNER_FOR_END28:%.*]]
// CHECK-32-EX:       omp.inner.for.body13:
// CHECK-32-EX-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP30:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP31:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-EX-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP31]] to i1
// CHECK-32-EX-NEXT:    [[FROMBOOL16:%.*]] = zext i1 [[TOBOOL14]] to i8
// CHECK-32-EX-NEXT:    store i8 [[FROMBOOL16]], ptr [[DOTCAPTURE_EXPR__CASTED15]], align 1
// CHECK-32-EX-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED15]], align 4
// CHECK-32-EX-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP34:%.*]] = inttoptr i32 [[TMP29]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP34]], ptr [[TMP33]], align 4
// CHECK-32-EX-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP36:%.*]] = inttoptr i32 [[TMP30]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP36]], ptr [[TMP35]], align 4
// CHECK-32-EX-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS17]], i32 0, i32 2
// CHECK-32-EX-NEXT:    [[TMP38:%.*]] = inttoptr i32 [[TMP32]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP38]], ptr [[TMP37]], align 4
// CHECK-32-EX-NEXT:    [[TMP39:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-EX-NEXT:    [[TOBOOL18:%.*]] = trunc i8 [[TMP39]] to i1
// CHECK-32-EX-NEXT:    [[TMP40:%.*]] = zext i1 [[TOBOOL18]] to i32
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP40]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined1, ptr null, ptr [[CAPTURED_VARS_ADDRS17]], i32 3)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC19:%.*]]
// CHECK-32-EX:       omp.inner.for.inc19:
// CHECK-32-EX-NEXT:    [[TMP41:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP42:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP41]], [[TMP42]]
// CHECK-32-EX-NEXT:    store i32 [[ADD20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP43:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP44:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP43]], [[TMP44]]
// CHECK-32-EX-NEXT:    store i32 [[ADD21]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP45:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP46:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD22:%.*]] = add nsw i32 [[TMP45]], [[TMP46]]
// CHECK-32-EX-NEXT:    store i32 [[ADD22]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP47:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP23:%.*]] = icmp sgt i32 [[TMP47]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP23]], label [[COND_TRUE24:%.*]], label [[COND_FALSE25:%.*]]
// CHECK-32-EX:       cond.true24:
// CHECK-32-EX-NEXT:    br label [[COND_END26:%.*]]
// CHECK-32-EX:       cond.false25:
// CHECK-32-EX-NEXT:    [[TMP48:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END26]]
// CHECK-32-EX:       cond.end26:
// CHECK-32-EX-NEXT:    [[COND27:%.*]] = phi i32 [ 9, [[COND_TRUE24]] ], [ [[TMP48]], [[COND_FALSE25]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND27]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP49:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP49]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND11]], !llvm.loop [[LOOP225:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end28:
// CHECK-32-EX-NEXT:    br label [[OMP_IF_END]]
// CHECK-32-EX:       omp_if.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP50:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP51:%.*]] = icmp ne i32 [[TMP50]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP51]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-32-EX:       omp_if.then:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP227]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP228:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-32-EX:       omp_if.else:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP12]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND2:%.*]]
// CHECK-32-EX:       omp.inner.for.cond2:
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[CMP3:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY4:%.*]], label [[OMP_INNER_FOR_END10:%.*]]
// CHECK-32-EX:       omp.inner.for.body4:
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK-32-EX-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
// CHECK-32-EX-NEXT:    store i32 [[ADD6]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE7:%.*]]
// CHECK-32-EX:       omp.body.continue7:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC8:%.*]]
// CHECK-32-EX:       omp.inner.for.inc8:
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK-32-EX-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND2]], !llvm.loop [[LOOP230:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end10:
// CHECK-32-EX-NEXT:    br label [[OMP_IF_END]]
// CHECK-32-EX:       omp_if.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP20]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l15_omp_outlined_omp_outlined1
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK-32-EX:       omp_if.then:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP4]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP231]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP232:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK-32-EX:       omp_if.else:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP12]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND2:%.*]]
// CHECK-32-EX:       omp.inner.for.cond2:
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[CMP3:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY4:%.*]], label [[OMP_INNER_FOR_END10:%.*]]
// CHECK-32-EX:       omp.inner.for.body4:
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK-32-EX-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
// CHECK-32-EX-NEXT:    store i32 [[ADD6]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE7:%.*]]
// CHECK-32-EX:       omp.body.continue7:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC8:%.*]]
// CHECK-32-EX:       omp.inner.for.inc8:
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// CHECK-32-EX-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND2]], !llvm.loop [[LOOP234:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end10:
// CHECK-32-EX-NEXT:    br label [[OMP_IF_END]]
// CHECK-32-EX:       omp_if.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP20]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP235]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP236:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l18_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP238]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP239:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP241]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP242:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l21_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP244]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP245:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP247]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP248:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l24_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP250]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP251:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP253]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP254:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l27_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP256]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP257:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP259]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP260:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l30_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP262]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP263:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP265]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP266:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l33_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP268]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP269:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[A:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP2]], ptr [[A_CASTED]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A_CASTED]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i32 [[TMP3]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [3 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[A1:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i32 4)
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A1]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP8]], ptr [[A_CASTED]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[A_CASTED]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = inttoptr i32 [[TMP9]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 3)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP22]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK-32-EX:       cond.true6:
// CHECK-32-EX-NEXT:    br label [[COND_END8:%.*]]
// CHECK-32-EX:       cond.false7:
// CHECK-32-EX-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END8]]
// CHECK-32-EX:       cond.end8:
// CHECK-32-EX-NEXT:    [[COND9:%.*]] = phi i32 [ 9, [[COND_TRUE6]] ], [ [[TMP23]], [[COND_FALSE7]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND9]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP24]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP26]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
// CHECK-32-EX:       .omp.lastprivate.then:
// CHECK-32-EX-NEXT:    [[TMP27:%.*]] = load i32, ptr [[A1]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP27]], ptr [[A_ADDR]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
// CHECK-32-EX:       .omp.lastprivate.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_free_shared(ptr [[A1]], i32 4)
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l37_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[A1:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP8]], ptr [[A1]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP12]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
// CHECK-32-EX:       .omp.lastprivate.then:
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[A1]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP13]], ptr [[A_ADDR]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
// CHECK-32-EX:       .omp.lastprivate.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l40_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l43_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l46_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP271]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP272:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l49_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP274]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP275:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l52_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP277]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP278:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l55_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP280]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP281:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP283]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP284:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l58_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP286]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP287:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[B:%.*]] = alloca [3 x i32], align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[B]], ptr align 4 @"__const.<captured>.b", i32 12, i1 false)
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP289]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP290:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l66_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP292]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP293:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[C:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i32 4)
// CHECK-32-EX-NEXT:    [[B:%.*]] = call align 8 ptr @__kmpc_alloc_shared(i32 4)
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP10]], ptr [[TMP9]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP8]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP12]], ptr [[TMP11]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP295]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP296:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    store ptr [[B]], ptr [[C]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_free_shared(ptr [[B]], i32 4)
// CHECK-32-EX-NEXT:    call void @__kmpc_free_shared(ptr [[C]], i32 4)
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP298]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP299:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP11]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined_wrapper
// CHECK-32-EX-SAME: (i16 noundef zeroext [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR7:[0-9]+]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTADDR:%.*]] = alloca i16, align 2
// CHECK-32-EX-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[GLOBAL_ARGS:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    store i16 [[TMP0]], ptr [[DOTADDR]], align 2
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = getelementptr inbounds ptr, ptr [[TMP2]], i32 0
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = getelementptr inbounds ptr, ptr [[TMP2]], i32 1
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l73_omp_outlined_omp_outlined(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], i32 [[TMP4]], i32 [[TMP6]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP301]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP302:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l81_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP304]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP305:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP307]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP308:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l85_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP310]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP311:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP313]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP314:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l89_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP316]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP317:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319:![0-9]+]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2), !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP319]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP320:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l93_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP322]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP323:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l97_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l101_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l105_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l109_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP325]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP326:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l113_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP328]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP329:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l117_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP331]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP332:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l121_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP334]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP335:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l125_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l130_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l135_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP3]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP3]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l140_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741862, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP337]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP338:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l145_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741861, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP340]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP341:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l150_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741859, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP343]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP344:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4
// CHECK-32-EX-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR2]]
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2]], i32 [[TMP1]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP5]], 10
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = inttoptr i32 [[TMP6]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP9]], ptr [[TMP8]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = inttoptr i32 [[TMP7]] to ptr
// CHECK-32-EX-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 2)
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
// CHECK-32-EX:       cond.true5:
// CHECK-32-EX-NEXT:    br label [[COND_END7:%.*]]
// CHECK-32-EX:       cond.false6:
// CHECK-32-EX-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END7]]
// CHECK-32-EX:       cond.end7:
// CHECK-32-EX-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND8]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l155_omp_outlined_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP5]], i32 1073741860, i32 [[TMP2]], i32 [[TMP3]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP5]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP346]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP347:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP5]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR8:[0-9]+]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i32
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP3]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l160_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-EX-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l163_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l166_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-EX-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l169_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP349]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP350:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l172_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP352]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP353:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l175_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP355]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP356:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l178_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP358]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP359:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i32
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 [[TMP3]], i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l181_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-EX-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP361]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP362:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-EX-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4:[0-9]+]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l185_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP364]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP365:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l189_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-EX-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP367]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP368:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-EX-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l193_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP370]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP371:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l197_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP373]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP374:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l201_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP376]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP377:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l205_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP379]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP380:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10:[0-9]+]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l209_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 65, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP382]]
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_fini_4(ptr @[[GLOB1]], i32 [[TMP1]]), !llvm.access.group [[ACC_GRP382]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP383:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l214_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP385]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP386:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l219_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-EX-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP388]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP389:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-EX-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP16]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l224_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP391]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP392:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l229_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP394]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP395:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l234_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP397]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP398:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR10]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l239_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP400]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP401:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
// CHECK-32-EX-NEXT:    br i1 [[TMP9]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK-32-EX:       .omp.final.then:
// CHECK-32-EX-NEXT:    store i32 10, ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK-32-EX:       .omp.final.done:
// CHECK-32-EX-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l244_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-EX-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l248_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK-32-EX:       omp.loop.exit:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l252_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP1]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK-32-EX:       cond.true:
// CHECK-32-EX-NEXT:    br label [[COND_END:%.*]]
// CHECK-32-EX:       cond.false:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[COND_END]]
// CHECK-32-EX:       cond.end:
// CHECK-32-EX-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK-32-EX-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-EX-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK-32-EX-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK-32-EX-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK-32-EX-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l256_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741862, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP403]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP404:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l260_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741861, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP406]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP407:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l264_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741859, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP409]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP410:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268
// CHECK-32-EX-SAME: (ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR8]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 4
// CHECK-32-EX-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_kernel_environment, ptr [[DYN_PTR]])
// CHECK-32-EX-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1
// CHECK-32-EX-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// CHECK-32-EX:       user_code.entry:
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK-32-EX-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP1]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 0)
// CHECK-32-EX-NEXT:    call void @__kmpc_target_deinit()
// CHECK-32-EX-NEXT:    ret void
// CHECK-32-EX:       worker.exit:
// CHECK-32-EX-NEXT:    ret void
//
//
// CHECK-32-EX-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l268_omp_outlined
// CHECK-32-EX-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// CHECK-32-EX-NEXT:  entry:
// CHECK-32-EX-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
// CHECK-32-EX-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK-32-EX-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK-32-EX-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK-32-EX-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 1073741860, i32 0, i32 9, i32 1, i32 1)
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK-32-EX:       omp.dispatch.cond:
// CHECK-32-EX-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_dispatch_next_4(ptr @[[GLOB1]], i32 [[TMP1]], ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]])
// CHECK-32-EX-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
// CHECK-32-EX-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK-32-EX:       omp.dispatch.body:
// CHECK-32-EX-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK-32-EX-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-32-EX:       omp.inner.for.cond:
// CHECK-32-EX-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412:![0-9]+]]
// CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32-EX:       omp.inner.for.body:
// CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-EX-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
// CHECK-32-EX-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK-32-EX-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-EX-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK-32-EX:       omp.body.continue:
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK-32-EX:       omp.inner.for.inc:
// CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-EX-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK-32-EX-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP412]]
// CHECK-32-EX-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP413:![0-9]+]]
// CHECK-32-EX:       omp.inner.for.end:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK-32-EX:       omp.dispatch.inc:
// CHECK-32-EX-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK-32-EX:       omp.dispatch.end:
// CHECK-32-EX-NEXT:    call void @__kmpc_dispatch_deinit(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK-32-EX-NEXT:    ret void
//