xref: /llvm-project/clang/test/OpenMP/target_parallel_generic_loop_codegen.cpp (revision b8cbc5c02c0c2da1077f125ad1c0f8643cdf8072)

// 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 _
// REQUIRES: amdgpu-registered-target

// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=IR-GPU

// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -emit-llvm %s -o - | FileCheck %s --check-prefix=IR

// Check same results after serialization round-trip
// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -emit-pch -o %t %s
// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR-PCH

// expected-no-diagnostics

#ifndef HEADER
#define HEADER

typedef void **omp_allocator_handle_t;
extern const omp_allocator_handle_t omp_null_allocator;
extern const omp_allocator_handle_t omp_default_mem_alloc;
extern const omp_allocator_handle_t omp_large_cap_mem_alloc;
extern const omp_allocator_handle_t omp_const_mem_alloc;
extern const omp_allocator_handle_t omp_high_bw_mem_alloc;
extern const omp_allocator_handle_t omp_low_lat_mem_alloc;
extern const omp_allocator_handle_t omp_cgroup_mem_alloc;
extern const omp_allocator_handle_t omp_pteam_mem_alloc;
extern const omp_allocator_handle_t omp_thread_mem_alloc;

extern int omp_get_thread_num(void);

#define N 64

int main() {
  int x = 0;
  int device_result[N] = {0};

  #pragma omp target parallel loop num_threads(N) uses_allocators(omp_pteam_mem_alloc) allocate(omp_pteam_mem_alloc: x) private(x) map(from: device_result)
  for (int i = 0; i < N; i++) {
    x = omp_get_thread_num();
    device_result[i] = i + x;
  }
}
#endif
// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37
// IR-GPU-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(256) [[DEVICE_RESULT:%.*]], ptr noundef [[OMP_PTEAM_MEM_ALLOC:%.*]]) #[[ATTR0:[0-9]+]] {
// IR-GPU-NEXT:  entry:
// IR-GPU-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// IR-GPU-NEXT:    [[DEVICE_RESULT_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// IR-GPU-NEXT:    [[OMP_PTEAM_MEM_ALLOC_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// IR-GPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [2 x ptr], align 8, addrspace(5)
// IR-GPU-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
// IR-GPU-NEXT:    [[DEVICE_RESULT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DEVICE_RESULT_ADDR]] to ptr
// IR-GPU-NEXT:    [[OMP_PTEAM_MEM_ALLOC_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[OMP_PTEAM_MEM_ALLOC_ADDR]] to ptr
// IR-GPU-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
// IR-GPU-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    store ptr [[DEVICE_RESULT]], ptr [[DEVICE_RESULT_ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    store ptr [[OMP_PTEAM_MEM_ALLOC]], ptr [[OMP_PTEAM_MEM_ALLOC_ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DEVICE_RESULT_ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37_kernel_environment to ptr), ptr [[DYN_PTR]])
// IR-GPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
// IR-GPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
// IR-GPU:       user_code.entry:
// IR-GPU-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1:[0-9]+]] to ptr))
// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[OMP_PTEAM_MEM_ALLOC_ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
// IR-GPU-NEXT:    store ptr [[TMP0]], ptr [[TMP4]], align 8
// IR-GPU-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
// IR-GPU-NEXT:    store ptr [[TMP3]], ptr [[TMP5]], align 8
// IR-GPU-NEXT:    call void @__kmpc_parallel_51(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP2]], i32 1, i32 64, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 2)
// IR-GPU-NEXT:    call void @__kmpc_target_deinit()
// IR-GPU-NEXT:    ret void
// IR-GPU:       worker.exit:
// IR-GPU-NEXT:    ret void
//
//
// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37_omp_outlined
// IR-GPU-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(256) [[DEVICE_RESULT:%.*]], ptr noundef [[OMP_PTEAM_MEM_ALLOC:%.*]]) #[[ATTR1:[0-9]+]] {
// IR-GPU-NEXT:  entry:
// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// IR-GPU-NEXT:    [[DEVICE_RESULT_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// IR-GPU-NEXT:    [[OMP_PTEAM_MEM_ALLOC_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// IR-GPU-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
// IR-GPU-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
// IR-GPU-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
// IR-GPU-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
// IR-GPU-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
// IR-GPU-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
// IR-GPU-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
// IR-GPU-NEXT:    [[DEVICE_RESULT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DEVICE_RESULT_ADDR]] to ptr
// IR-GPU-NEXT:    [[OMP_PTEAM_MEM_ALLOC_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[OMP_PTEAM_MEM_ALLOC_ADDR]] to ptr
// IR-GPU-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
// IR-GPU-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
// IR-GPU-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
// IR-GPU-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
// IR-GPU-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
// IR-GPU-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
// IR-GPU-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
// IR-GPU-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    store ptr [[DEVICE_RESULT]], ptr [[DEVICE_RESULT_ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    store ptr [[OMP_PTEAM_MEM_ALLOC]], ptr [[OMP_PTEAM_MEM_ALLOC_ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DEVICE_RESULT_ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
// IR-GPU-NEXT:    store i32 63, ptr [[DOTOMP_UB_ASCAST]], align 4
// IR-GPU-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
// IR-GPU-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// IR-GPU-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2:[0-9]+]] to ptr), i32 [[TMP2]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
// IR-GPU-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// IR-GPU:       omp.dispatch.cond:
// IR-GPU-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// IR-GPU-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 63
// IR-GPU-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// IR-GPU:       cond.true:
// IR-GPU-NEXT:    br label [[COND_END:%.*]]
// IR-GPU:       cond.false:
// IR-GPU-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// IR-GPU-NEXT:    br label [[COND_END]]
// IR-GPU:       cond.end:
// IR-GPU-NEXT:    [[COND:%.*]] = phi i32 [ 63, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// IR-GPU-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// IR-GPU-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// IR-GPU-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// IR-GPU-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// IR-GPU:       omp.dispatch.body:
// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// IR-GPU:       omp.inner.for.cond:
// IR-GPU-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// IR-GPU-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
// IR-GPU-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// IR-GPU:       omp.inner.for.body:
// IR-GPU-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// IR-GPU-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
// IR-GPU-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// IR-GPU-NEXT:    store i32 [[ADD]], ptr [[I_ASCAST]], align 4
// IR-GPU-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z18omp_get_thread_numv() #[[ATTR5:[0-9]+]]
// IR-GPU-NEXT:    store i32 [[CALL]], ptr addrspacecast (ptr addrspace(3) @x to ptr), align 4
// IR-GPU-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP12:%.*]] = load i32, ptr addrspacecast (ptr addrspace(3) @x to ptr), align 4
// IR-GPU-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// IR-GPU-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I_ASCAST]], align 4
// IR-GPU-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
// IR-GPU-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [64 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// IR-GPU-NEXT:    store i32 [[ADD3]], ptr [[ARRAYIDX]], align 4
// IR-GPU-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// IR-GPU:       omp.body.continue:
// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// IR-GPU:       omp.inner.for.inc:
// IR-GPU-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
// IR-GPU-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], 1
// IR-GPU-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV_ASCAST]], align 4
// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND]]
// IR-GPU:       omp.inner.for.end:
// IR-GPU-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// IR-GPU:       omp.dispatch.inc:
// IR-GPU-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// IR-GPU-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
// IR-GPU-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_LB_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB_ASCAST]], align 4
// IR-GPU-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
// IR-GPU-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
// IR-GPU-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_UB_ASCAST]], align 4
// IR-GPU-NEXT:    br label [[OMP_DISPATCH_COND]]
// IR-GPU:       omp.dispatch.end:
// IR-GPU-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP2]])
// IR-GPU-NEXT:    ret void
//
//
// IR-LABEL: define {{[^@]+}}@main
// IR-SAME: () #[[ATTR0:[0-9]+]] {
// IR-NEXT:  entry:
// IR-NEXT:    [[X:%.*]] = alloca i32, align 4
// IR-NEXT:    [[DEVICE_RESULT:%.*]] = alloca [64 x i32], align 16
// IR-NEXT:    store i32 0, ptr [[X]], align 4
// IR-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[DEVICE_RESULT]], i8 0, i64 256, i1 false)
// IR-NEXT:    [[TMP0:%.*]] = load ptr, ptr @omp_pteam_mem_alloc, align 8
// IR-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37(ptr [[DEVICE_RESULT]], ptr [[TMP0]]) #[[ATTR3:[0-9]+]]
// IR-NEXT:    ret i32 0
//
//
// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37
// IR-SAME: (ptr noundef nonnull align 4 dereferenceable(256) [[DEVICE_RESULT:%.*]], ptr noundef [[OMP_PTEAM_MEM_ALLOC:%.*]]) #[[ATTR2:[0-9]+]] {
// IR-NEXT:  entry:
// IR-NEXT:    [[DEVICE_RESULT_ADDR:%.*]] = alloca ptr, align 8
// IR-NEXT:    [[OMP_PTEAM_MEM_ALLOC_ADDR:%.*]] = alloca ptr, align 8
// IR-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2:[0-9]+]])
// IR-NEXT:    store ptr [[DEVICE_RESULT]], ptr [[DEVICE_RESULT_ADDR]], align 8
// IR-NEXT:    store ptr [[OMP_PTEAM_MEM_ALLOC]], ptr [[OMP_PTEAM_MEM_ALLOC_ADDR]], align 8
// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DEVICE_RESULT_ADDR]], align 8
// IR-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB2]], i32 [[TMP0]], i32 64)
// IR-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[OMP_PTEAM_MEM_ALLOC_ADDR]], align 8
// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.omp_outlined, ptr [[TMP1]], ptr [[TMP2]])
// IR-NEXT:    ret void
//
//
// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.omp_outlined
// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(256) [[DEVICE_RESULT:%.*]], ptr noundef [[OMP_PTEAM_MEM_ALLOC:%.*]]) #[[ATTR2]] {
// IR-NEXT:  entry:
// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// IR-NEXT:    [[DEVICE_RESULT_ADDR:%.*]] = alloca ptr, align 8
// IR-NEXT:    [[OMP_PTEAM_MEM_ALLOC_ADDR:%.*]] = alloca ptr, align 8
// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// IR-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// IR-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// IR-NEXT:    store ptr [[DEVICE_RESULT]], ptr [[DEVICE_RESULT_ADDR]], align 8
// IR-NEXT:    store ptr [[OMP_PTEAM_MEM_ALLOC]], ptr [[OMP_PTEAM_MEM_ALLOC_ADDR]], align 8
// IR-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DEVICE_RESULT_ADDR]], align 8
// IR-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// IR-NEXT:    store i32 63, ptr [[DOTOMP_UB]], align 4
// IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// IR-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr @omp_pteam_mem_alloc, align 8
// IR-NEXT:    [[DOTX__VOID_ADDR:%.*]] = call ptr @__kmpc_alloc(i32 [[TMP2]], i64 4, ptr [[TMP3]])
// IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// IR-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 63
// IR-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// IR:       cond.true:
// IR-NEXT:    br label [[COND_END:%.*]]
// IR:       cond.false:
// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// IR-NEXT:    br label [[COND_END]]
// IR:       cond.end:
// IR-NEXT:    [[COND:%.*]] = phi i32 [ 63, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// IR-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// IR:       omp.inner.for.cond:
// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// IR-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// IR-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// IR-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// IR:       omp.inner.for.cond.cleanup:
// IR-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// IR:       omp.inner.for.body:
// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// IR-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// IR-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// IR-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// IR-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z18omp_get_thread_numv()
// IR-NEXT:    store i32 [[CALL]], ptr [[DOTX__VOID_ADDR]], align 4
// IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTX__VOID_ADDR]], align 4
// IR-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
// IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [64 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// IR-NEXT:    store i32 [[ADD2]], ptr [[ARRAYIDX]], align 4
// IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// IR:       omp.body.continue:
// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// IR:       omp.inner.for.inc:
// IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// IR-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
// IR-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
// IR:       omp.inner.for.end:
// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// IR:       omp.loop.exit:
// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
// IR-NEXT:    [[TMP14:%.*]] = load ptr, ptr @omp_pteam_mem_alloc, align 8
// IR-NEXT:    call void @__kmpc_free(i32 [[TMP2]], ptr [[DOTX__VOID_ADDR]], ptr [[TMP14]])
// IR-NEXT:    ret void
//
//
// IR-PCH-LABEL: define {{[^@]+}}@main
// IR-PCH-SAME: () #[[ATTR0:[0-9]+]] {
// IR-PCH-NEXT:  entry:
// IR-PCH-NEXT:    [[X:%.*]] = alloca i32, align 4
// IR-PCH-NEXT:    [[DEVICE_RESULT:%.*]] = alloca [64 x i32], align 16
// IR-PCH-NEXT:    store i32 0, ptr [[X]], align 4
// IR-PCH-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[DEVICE_RESULT]], i8 0, i64 256, i1 false)
// IR-PCH-NEXT:    [[TMP0:%.*]] = load ptr, ptr @omp_pteam_mem_alloc, align 8
// IR-PCH-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37(ptr [[DEVICE_RESULT]], ptr [[TMP0]]) #[[ATTR3:[0-9]+]]
// IR-PCH-NEXT:    ret i32 0
//
//
// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37
// IR-PCH-SAME: (ptr noundef nonnull align 4 dereferenceable(256) [[DEVICE_RESULT:%.*]], ptr noundef [[OMP_PTEAM_MEM_ALLOC:%.*]]) #[[ATTR2:[0-9]+]] {
// IR-PCH-NEXT:  entry:
// IR-PCH-NEXT:    [[DEVICE_RESULT_ADDR:%.*]] = alloca ptr, align 8
// IR-PCH-NEXT:    [[OMP_PTEAM_MEM_ALLOC_ADDR:%.*]] = alloca ptr, align 8
// IR-PCH-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2:[0-9]+]])
// IR-PCH-NEXT:    store ptr [[DEVICE_RESULT]], ptr [[DEVICE_RESULT_ADDR]], align 8
// IR-PCH-NEXT:    store ptr [[OMP_PTEAM_MEM_ALLOC]], ptr [[OMP_PTEAM_MEM_ALLOC_ADDR]], align 8
// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DEVICE_RESULT_ADDR]], align 8
// IR-PCH-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB2]], i32 [[TMP0]], i32 64)
// IR-PCH-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[OMP_PTEAM_MEM_ALLOC_ADDR]], align 8
// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.omp_outlined, ptr [[TMP1]], ptr [[TMP2]])
// IR-PCH-NEXT:    ret void
//
//
// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.omp_outlined
// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(256) [[DEVICE_RESULT:%.*]], ptr noundef [[OMP_PTEAM_MEM_ALLOC:%.*]]) #[[ATTR2]] {
// IR-PCH-NEXT:  entry:
// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// IR-PCH-NEXT:    [[DEVICE_RESULT_ADDR:%.*]] = alloca ptr, align 8
// IR-PCH-NEXT:    [[OMP_PTEAM_MEM_ALLOC_ADDR:%.*]] = alloca ptr, align 8
// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// IR-PCH-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// IR-PCH-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// IR-PCH-NEXT:    store ptr [[DEVICE_RESULT]], ptr [[DEVICE_RESULT_ADDR]], align 8
// IR-PCH-NEXT:    store ptr [[OMP_PTEAM_MEM_ALLOC]], ptr [[OMP_PTEAM_MEM_ALLOC_ADDR]], align 8
// IR-PCH-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DEVICE_RESULT_ADDR]], align 8
// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// IR-PCH-NEXT:    store i32 63, ptr [[DOTOMP_UB]], align 4
// IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// IR-PCH-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr @omp_pteam_mem_alloc, align 8
// IR-PCH-NEXT:    [[DOTX__VOID_ADDR:%.*]] = call ptr @__kmpc_alloc(i32 [[TMP2]], i64 4, ptr [[TMP3]])
// IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// IR-PCH-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 63
// IR-PCH-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// IR-PCH:       cond.true:
// IR-PCH-NEXT:    br label [[COND_END:%.*]]
// IR-PCH:       cond.false:
// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// IR-PCH-NEXT:    br label [[COND_END]]
// IR-PCH:       cond.end:
// IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ 63, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// IR-PCH-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// IR-PCH:       omp.inner.for.cond:
// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// IR-PCH-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// IR-PCH-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// IR-PCH-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// IR-PCH:       omp.inner.for.cond.cleanup:
// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// IR-PCH:       omp.inner.for.body:
// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
// IR-PCH-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z18omp_get_thread_numv()
// IR-PCH-NEXT:    store i32 [[CALL]], ptr [[DOTX__VOID_ADDR]], align 4
// IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTX__VOID_ADDR]], align 4
// IR-PCH-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
// IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
// IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [64 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// IR-PCH-NEXT:    store i32 [[ADD2]], ptr [[ARRAYIDX]], align 4
// IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// IR-PCH:       omp.body.continue:
// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// IR-PCH:       omp.inner.for.inc:
// IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// IR-PCH-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
// IR-PCH-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
// IR-PCH:       omp.inner.for.end:
// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// IR-PCH:       omp.loop.exit:
// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
// IR-PCH-NEXT:    [[TMP14:%.*]] = load ptr, ptr @omp_pteam_mem_alloc, align 8
// IR-PCH-NEXT:    call void @__kmpc_free(i32 [[TMP2]], ptr [[DOTX__VOID_ADDR]], ptr [[TMP14]])
// IR-PCH-NEXT:    ret void
//