starpu-max/mpi/tests/checkpoints.c

/* StarPU --- Runtime system for heterogeneous multicore architectures.
 *
 * Copyright (C) 2013-2020  Université de Bordeaux, CNRS (LaBRI UMR 5800), Inria
 *
 * 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_mpi.h>
#include "helper.h"

#include <mpi_failure_tolerance/starpu_mpi_checkpoint_template.h>

#define ARRAY_SIZE 12

#define STARPU_MPI_INIT(void) do{struct starpu_conf conf; int ret; \
starpu_conf_init(&conf); \
conf.nmic = 0; \
conf.nmpi_ms = 0; \
ret = starpu_init(NULL); \
if (STARPU_UNLIKELY(ret == -ENODEV)) \
{ \
return 77; \
} \
STARPU_CHECK_RETURN_VALUE(ret, "starpu_init"); \
if (starpu_cpu_worker_get_count() < 1) \
{ \
FPRINTF(stderr, "This application requires at least 1 cpu worker\n"); \
starpu_shutdown(); \
return 77; \
} \
starpu_mpi_init(&argc, &argv, 1); \
starpu_mpi_comm_size(MPI_COMM_WORLD, &nb_nodes); \
starpu_mpi_comm_rank(MPI_COMM_WORLD, &me); \
}while(0)


int nb_nodes;
int me;

int backup_of(int _me)
{
	return (_me+1)%nb_nodes;
}

int backuped_by(int _me)
{
	return (_me-1)%nb_nodes;
}

int pseudotest_checkpoint_template_register(int argc, char* argv[])
{
	starpu_data_handle_t             h;
	starpu_data_handle_t             h_array[ARRAY_SIZE];
	starpu_mpi_checkpoint_template_t cp_template1, cp_template2;
	int                              val = 42;
	int                              val2 = 1234;

	int array[ARRAY_SIZE];
	int ret;

	//init array
	for (int i=0 ; i<ARRAY_SIZE ; i++)
	{
		array[i] = i*1111+42;
	}

	for (int i=0 ; i<ARRAY_SIZE ; i++)
	{
		h_array[i] = NULL;
	}
	FPRINTF(stderr, "Go\n");

	STARPU_MPI_INIT();

	FPRINTF_MPI(stderr, "Init ok - my rnk %d - size %d\n", me, nb_nodes);

	starpu_variable_data_register(&h, STARPU_MAIN_RAM, (uintptr_t)&val2, sizeof(int));

	starpu_vector_data_register(h_array, STARPU_MAIN_RAM, (uintptr_t)array, ARRAY_SIZE, sizeof(int));
	for (int i=0 ; i<ARRAY_SIZE ; i++)
	{
		starpu_variable_data_register(&h_array[i], STARPU_MAIN_RAM, (uintptr_t)&array[i], sizeof(int));
	}

	starpu_mpi_checkpoint_template_register(&cp_template1, 123486,
	                                        STARPU_VALUE, &val, sizeof(int), 84, backup_of,
	                                        STARPU_R, &h, 1,
	                                        0);

	FPRINTF(stderr, "registered!\n");
	_starpu_mpi_checkpoint_template_print(cp_template1);

	starpu_mpi_checkpoint_template_create(&cp_template2, 98765);
	starpu_mpi_checkpoint_template_add_entry(&cp_template2, STARPU_R, &h, 1);
	starpu_mpi_checkpoint_template_add_entry(&cp_template2, STARPU_VALUE, &val, sizeof(int), 84, backup_of);
	starpu_mpi_checkpoint_template_freeze(&cp_template2);

	FPRINTF(stderr, "registered 2!\n");
	_starpu_mpi_checkpoint_template_print(cp_template1);

	starpu_shutdown();
	return 0;
}

int test_checkpoint_submit(int argc, char* argv[])
{
	starpu_data_handle_t handle0, handle1;
	starpu_mpi_checkpoint_template_t cp_template;
	int val0 = 0;
	int val1 = 0;

	FPRINTF(stderr, "Go\n");

	STARPU_MPI_INIT();

	FPRINTF_MPI(stderr, "Init ok - my rnk %d - size %d\n", me, nb_nodes);

	starpu_variable_data_register(&handle0, STARPU_MAIN_RAM, (uintptr_t)&val0, sizeof(int));
	starpu_variable_data_register(&handle1, STARPU_MAIN_RAM, (uintptr_t)&val1, sizeof(int));

	starpu_mpi_data_register(handle0, 100, 0);
	starpu_mpi_data_register(handle1, 200, 1);

	starpu_mpi_checkpoint_template_register(&cp_template, 321,
			STARPU_R, handle0, 1,
			STARPU_R, handle1, 0,
			0);

	switch (me)
	{
		case 0:
			val0 = 42;
			break;
		case 1:
			val1 = 1000;
			break;
		default:
			STARPU_ABORT_MSG("Test code with only two node communicator.\n");
			break;
	}
	FPRINTF_MPI(stderr, "Submitting\n");
	starpu_mpi_submit_checkpoint_template(cp_template);

	FPRINTF_MPI(stderr, "Submitted\n");
	sleep(10);
	FPRINTF_MPI(stderr, "Bye!\n");
	starpu_shutdown();

	return 0;
}

int main(int argc, char* argv[])
{
	pseudotest_checkpoint_template_register(argc, argv);
	//test_checkpoint_submit(argc, argv);
	return 0;
}