starpu-max/tests/overlap/overlap.c

/*
 * StarPU
 * Copyright (C) INRIA 2008-2009 (see AUTHORS file)
 *
 * This program 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.
 *
 * This program 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 <stdio.h>
#include <unistd.h>
#include <errno.h>
#include <starpu.h>
#include <stdlib.h>

#define NTASKS	10000
#define VECTORSIZE	1024
#define TASKDURATION	24U

#define SYMBOL	"sleep"

static pthread_mutex_t mutex;
static pthread_cond_t cond;

static unsigned finished = 0;
static unsigned cnt = NTASKS;

starpu_data_handle handle;
static float *buffer;

static void callback(void *arg)
{
	unsigned res = STARPU_ATOMIC_ADD(&cnt, -1);

	if (res == 0)
	{
		pthread_mutex_lock(&mutex);
		finished = 1;
		pthread_cond_signal(&cond);
		pthread_mutex_unlock(&mutex);
	}
}

static void codelet_sleep(void *descr[], __attribute__ ((unused)) void *_args)
{
	usleep(TASKDURATION);
}

static struct starpu_perfmodel_t model = {
	.type = STARPU_HISTORY_BASED,
	.symbol = NULL /* to be defined later */
};

static starpu_codelet cl = {
	.where = STARPU_CPU|STARPU_CUDA,
	.cpu_func = codelet_sleep,
	.cuda_func = codelet_sleep,
	.nbuffers = 1,
	.model =  &model
};

static char symbolname[128];

int main(int argc, char **argv)
{
	unsigned i;

	starpu_init(NULL);

	/* create data */
	starpu_malloc_pinned_if_possible((void **)&buffer, NTASKS*VECTORSIZE*sizeof(char));

	/* declare data to StarPU */
	starpu_register_vector_data(&handle, 0, (uintptr_t)buffer,
					NTASKS*VECTORSIZE, sizeof(char));

	starpu_filter f =
	{
		.filter_func = starpu_block_filter_func_vector,
		.filter_arg = NTASKS
	};

	starpu_partition_data(handle, &f);

	snprintf(symbolname, 128, "overlap_sleep_%d_%d", VECTORSIZE, TASKDURATION);

	model.symbol = symbolname;

	unsigned iter;
	for (iter = 0; iter < NTASKS; iter++)
	{
		struct starpu_task *task = starpu_task_create();
		task->cl = &cl;

		task->buffers[0].handle = starpu_get_sub_data(handle, 1, iter);
		task->buffers[0].mode = STARPU_R;

		task->callback_func = callback;
		task->callback_arg = NULL;

		int ret = starpu_submit_task(task);
		if (ret == -ENODEV)
			goto enodev;
	}

	pthread_mutex_lock(&mutex);
	if (!finished)
		pthread_cond_wait(&cond, &mutex);
	pthread_mutex_unlock(&mutex);

	starpu_shutdown();

	return 0;

enodev:
	fprintf(stderr, "WARNING: No one can execute this task\n");
	/* yes, we do not perform the computation but we did detect that no one
 	 * could perform the kernel, so this is not an error from StarPU */
	return 0;
}