starpu-max/examples/sched_ctx/parallel_code.c

/* StarPU --- Runtime system for heterogeneous multicore architectures.
 *
 * Copyright (C) 2010-2013  Université de Bordeaux 1
 * Copyright (C) 2010-2013  Centre National de la Recherche Scientifique
 *
 * StarPU is free software; you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License as published by
 * the Free Software Foundation; either version 2.1 of the License, or (at
 * your option) any later version.
 *
 * StarPU is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 *
 * See the GNU Lesser General Public License in COPYING.LGPL for more details.
 */

#include <starpu.h>

#ifdef STARPU_QUICK_CHECK
#define NTASKS 64
#else
#define NTASKS 1000
#endif

int tasks_executed = 0;
starpu_pthread_mutex_t mut;

static void sched_ctx_func(void *descr[] STARPU_ATTRIBUTE_UNUSED, void *arg STARPU_ATTRIBUTE_UNUSED)
{
	starpu_pthread_mutex_lock(&mut);
	tasks_executed++;
	starpu_pthread_mutex_unlock(&mut);
}

static struct starpu_codelet sched_ctx_codelet =
{
	.cpu_funcs = {sched_ctx_func, NULL},
	.cuda_funcs = {sched_ctx_func, NULL},
	.opencl_funcs = {sched_ctx_func, NULL},
	.model = NULL,
	.nbuffers = 0,
	.name = "sched_ctx"
};

int parallel_code(int nprocs)
{
	int i;
	int tasks = 0;
#pragma omp parallel for num_threads(nprocs)
	for (i = 0; i < NTASKS; i++) 
		tasks++;

	return tasks;
}

int main(int argc, char **argv)
{
	int ntasks = NTASKS;
	int ret;

	ret = starpu_init(NULL);
	if (ret == -ENODEV)
		return 77;
	STARPU_CHECK_RETURN_VALUE(ret, "starpu_init");

	starpu_pthread_mutex_init(&mut, NULL);
	int nprocs1 = 1;
	int nprocs2 = 1;
	int procs1[20], procs2[20];
	procs1[0] = 0;
	procs2[0] = 0;

#ifdef STARPU_USE_CPU
	unsigned ncpus =  starpu_cpu_worker_get_count();
	starpu_worker_get_ids_by_type(STARPU_CPU_WORKER, procs1, ncpus);

	nprocs1 = ncpus/2;
	nprocs2 =  nprocs1;
	int j, k = 0;
	for(j = nprocs1; j < nprocs1+nprocs2; j++)
		procs2[k++] = j;
#endif

	/*create contexts however you want*/
	unsigned sched_ctx1 = starpu_sched_ctx_create(procs1, nprocs1, "ctx1", STARPU_SCHED_CTX_POLICY_NAME, "dmda", 0);
	unsigned sched_ctx2 = starpu_sched_ctx_create(procs2, nprocs2, "ctx2", STARPU_SCHED_CTX_POLICY_NAME, "dmda", 0);

	/*indicate what to do with the resources when context 2 finishes (it depends on your application)*/
	starpu_sched_ctx_set_inheritor(sched_ctx2, sched_ctx1);

	int i;
	for (i = 0; i < ntasks; i++)
	{
		struct starpu_task *task = starpu_task_create();

		task->cl = &sched_ctx_codelet;
		task->cl_arg = NULL;

		/*submit tasks to context*/
		ret = starpu_task_submit_to_ctx(task,sched_ctx1);

		STARPU_CHECK_RETURN_VALUE(ret, "starpu_task_submit");
	}

	/* tell starpu when you finished submitting tasks to this context
	   in order to allow moving resources from this context to the inheritor one
	   when its corresponding tasks finished executing */

	starpu_sched_ctx_finished_submit(sched_ctx1);

	/* execute an openmp code */
	int ret_ntasks = (int)starpu_sched_ctx_exec_parallel_code((void*)parallel_code, (void*)nprocs2, sched_ctx2);
	starpu_sched_ctx_finished_submit(sched_ctx2);

	/* wait for all tasks at the end*/
	starpu_task_wait_for_all();

	starpu_sched_ctx_delete(sched_ctx1);
	starpu_sched_ctx_delete(sched_ctx2);
	printf("ctx%d: tasks starpu executed %d out of %d\n", sched_ctx1, tasks_executed, ntasks);
	printf("ctx%d: tasks openmp executed %d out of %d\n", sched_ctx2, ret_ntasks, NTASKS);
	starpu_shutdown();

	return 0;
}