void shmem_parallel_for_nbi(void (*body)(void *), void *user_data, shmem_future_t **optional_future,
int lowBound, int highBound, int stride, int tile_size,
int loop_dimension, int loop_mode) {
loop_domain_t info = { lowBound, highBound, stride, tile_size};
hclib_forasync(body, user_data, optional_future, loop_dimension, &info, loop_mode);
}
hclib_future_t *hclib_forasync_future(void *forasync_fct, void *argv,
int dim, hclib_loop_domain_t *domain,
forasync_mode_t mode) {
hclib_start_finish();
hclib_forasync(forasync_fct, argv, dim, domain, mode);
return hclib_end_finish_nonblocking();
}
void entrypoint(void *arg) {
int *ran = (int *)arg;
// This is ok to have these on stack because this
// code is alive until the end of the program.
init_ran(ran, H1*H2);
loop_domain_t loop0 = {0,H1,1,T1};
loop_domain_t loop1 = {0,H2,1,T2};
loop_domain_t loop[2] = {loop0, loop1};
hclib_start_finish();
hclib_forasync(forasync_fct2, (void*)(ran), NULL, 2, loop,
FORASYNC_MODE_RECURSIVE);
hclib_end_finish();
printf("Call Finalize\n");
}