starpu-max/src/sched_policies/parallel_heft.c

/* StarPU --- Runtime system for heterogeneous multicore architectures.
 *
 * Copyright (C) 2010-2011  Université de Bordeaux 1
 *
 * 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.
 */

/* Distributed queues using performance modeling to assign tasks */

#include <float.h>
#include <limits.h>
#include <core/workers.h>
#include <core/perfmodel/perfmodel.h>
#include <starpu_parameters.h>
#include <common/barrier.h>

static pthread_mutex_t big_lock;

static unsigned nworkers, ncombinedworkers;
//static enum starpu_perf_archtype applicable_perf_archtypes[STARPU_NARCH_VARIATIONS];
//static unsigned napplicable_perf_archtypes = 0;

static pthread_cond_t sched_cond[STARPU_NMAXWORKERS];
static pthread_mutex_t sched_mutex[STARPU_NMAXWORKERS];

static double alpha = STARPU_DEFAULT_ALPHA;
static double beta = STARPU_DEFAULT_BETA;
static double _gamma = STARPU_DEFAULT_GAMMA;
static double idle_power = 0.0;

static double worker_exp_start[STARPU_NMAXWORKERS];
static double worker_exp_end[STARPU_NMAXWORKERS];
static double worker_exp_len[STARPU_NMAXWORKERS];
static int ntasks[STARPU_NMAXWORKERS];

static void parallel_heft_post_exec_hook(struct starpu_task *task)
{
	if (!task->cl || task->execute_on_a_specific_worker)
		return;

	int workerid = starpu_worker_get_id();
	double model = task->predicted;
	
	if (model < 0.0)
		model = 0.0;
	
	/* Once we have executed the task, we can update the predicted amount
	 * of work. */
	PTHREAD_MUTEX_LOCK(&sched_mutex[workerid]);
	worker_exp_len[workerid] -= model;
	worker_exp_start[workerid] = starpu_timing_now();
	worker_exp_end[workerid] = worker_exp_start[workerid] + worker_exp_len[workerid];
	ntasks[workerid]--;
	PTHREAD_MUTEX_UNLOCK(&sched_mutex[workerid]);
}

static int push_task_on_best_worker(struct starpu_task *task, int best_workerid, double exp_end_predicted, int prio)
{
	/* make sure someone coule execute that task ! */
	STARPU_ASSERT(best_workerid != -1);

	/* Is this a basic worker or a combined worker ? */
	int nbasic_workers = (int)starpu_worker_get_count();
	int is_basic_worker = (best_workerid < nbasic_workers);

	unsigned memory_node; 
	memory_node = starpu_worker_get_memory_node(best_workerid);

	if (starpu_get_prefetch_flag())
		starpu_prefetch_task_input_on_node(task, memory_node);

	int ret = 0;

	PTHREAD_MUTEX_LOCK(&big_lock);

	if (is_basic_worker)
	{
		task->predicted = exp_end_predicted - worker_exp_end[best_workerid];
		worker_exp_len[best_workerid] += exp_end_predicted - worker_exp_end[best_workerid];
		worker_exp_end[best_workerid] = exp_end_predicted;
		worker_exp_start[best_workerid] = exp_end_predicted - worker_exp_len[best_workerid];
	
		ntasks[best_workerid]++;

		ret = starpu_push_local_task(best_workerid, task, prio);
	}
	else {
		/* This is a combined worker so we create task aliases */
		struct starpu_combined_worker_s *combined_worker;
		combined_worker = _starpu_get_combined_worker_struct(best_workerid);
		int worker_size = combined_worker->worker_size;
		int *combined_workerid = combined_worker->combined_workerid;

		starpu_job_t j = _starpu_get_job_associated_to_task(task);
		j->task_size = worker_size;
		j->combined_workerid = best_workerid;
		j->active_task_alias_count = 0;

		PTHREAD_BARRIER_INIT(&j->before_work_barrier, NULL, worker_size);
		PTHREAD_BARRIER_INIT(&j->after_work_barrier, NULL, worker_size);

		int i;
		for (i = 0; i < worker_size; i++)
		{
			struct starpu_task *alias = _starpu_create_task_alias(task);
			int local_worker = combined_workerid[i];

			alias->predicted = exp_end_predicted - worker_exp_end[local_worker];
	
			worker_exp_len[local_worker] += exp_end_predicted - worker_exp_end[local_worker];
			worker_exp_end[local_worker] = exp_end_predicted;
			worker_exp_start[local_worker] = exp_end_predicted - worker_exp_len[local_worker];
		
			ntasks[local_worker]++;
	
			ret |= starpu_push_local_task(local_worker, alias, prio);
		}

	}

	PTHREAD_MUTEX_UNLOCK(&big_lock);

	return ret;
}

static double compute_expected_end(int workerid, double length)
{
	if (workerid < (int)nworkers)
	{
		/* This is a basic worker */
		return worker_exp_start[workerid] + worker_exp_len[workerid] + length;
	}
	else {
		/* This is a combined worker, the expected end is the end for the latest worker */
		int worker_size;
		int *combined_workerid;
		starpu_combined_worker_get_description(workerid, &worker_size, &combined_workerid);

		double exp_end = DBL_MIN;

		int i;
		for (i = 0; i < worker_size; i++)
		{
			double local_exp_start = worker_exp_start[combined_workerid[i]];
			double local_exp_len = worker_exp_len[combined_workerid[i]];
			double local_exp_end = local_exp_start + local_exp_len + length;
			exp_end = STARPU_MAX(exp_end, local_exp_end);
		}

		return exp_end;
	}
}

static double compute_ntasks_end(int workerid)
{
	enum starpu_perf_archtype perf_arch = starpu_worker_get_perf_archtype(workerid);
	if (workerid < (int)nworkers)
	{
		/* This is a basic worker */
		return ntasks[workerid] / starpu_worker_get_relative_speedup(perf_arch);
	}
	else {
		/* This is a combined worker, the expected end is the end for the latest worker */
		int worker_size;
		int *combined_workerid;
		starpu_combined_worker_get_description(workerid, &worker_size, &combined_workerid);

		int ntasks_end=0;

		int i;
		for (i = 0; i < worker_size; i++)
		{
			/* XXX: this is actually bogus: not all pushed tasks are necessarily parallel... */
			ntasks_end = STARPU_MAX(ntasks_end, ntasks[combined_workerid[i]] / starpu_worker_get_relative_speedup(perf_arch));
		}

		return ntasks_end;
	}
}

static int _parallel_heft_push_task(struct starpu_task *task, unsigned prio)
{
	unsigned worker;
	int best = -1;
	
	/* this flag is set if the corresponding worker is selected because
	   there is no performance prediction available yet */
	int forced_best = -1;

	double local_task_length[nworkers+ncombinedworkers];
	double local_data_penalty[nworkers+ncombinedworkers];
	double local_power[nworkers+ncombinedworkers];
	double local_exp_end[nworkers+ncombinedworkers];
	double fitness[nworkers+ncombinedworkers];

	double max_exp_end = 0.0;

	int skip_worker[nworkers+ncombinedworkers];

	double best_exp_end = DBL_MAX;
	//double penality_best = 0.0;

	int ntasks_best = -1;
	double ntasks_best_end = 0.0;
	int calibrating = 0;

	/* A priori, we know all estimations */
	int unknown = 0;

	for (worker = 0; worker < nworkers; worker++)
	{
		/* Sometimes workers didn't take the tasks as early as we expected */
		worker_exp_start[worker] = STARPU_MAX(worker_exp_start[worker], starpu_timing_now());
		worker_exp_end[worker] = worker_exp_start[worker] + worker_exp_len[worker];
		if (worker_exp_end[worker] > max_exp_end)
			max_exp_end = worker_exp_end[worker];
	}

	for (worker = 0; worker < (nworkers+ncombinedworkers); worker++)
	{
		if (!starpu_combined_worker_may_execute_task(worker, task))
		{
			/* no one on that queue may execute this task */
			skip_worker[worker] = 1;
			continue;
		}
		else {
			skip_worker[worker] = 0;
		}

		enum starpu_perf_archtype perf_arch = starpu_worker_get_perf_archtype(worker);
		local_task_length[worker] = starpu_task_expected_length(task, perf_arch);

		unsigned memory_node = starpu_worker_get_memory_node(worker);
		local_data_penalty[worker] = starpu_task_expected_data_transfer_time(memory_node, task);

		double ntasks_end = compute_ntasks_end(worker);

		if (ntasks_best == -1
				|| (!calibrating && ntasks_end < ntasks_best_end) /* Not calibrating, take better task */
				|| (!calibrating && local_task_length[worker] == -1.0) /* Not calibrating but this worker is being calibrated */
				|| (calibrating && local_task_length[worker] == -1.0 && ntasks_end < ntasks_best_end) /* Calibrating, compete this worker with other non-calibrated */
				) {
			ntasks_best_end = ntasks_end;
			ntasks_best = worker;
		}

		if (local_task_length[worker] == -1.0)
			/* we are calibrating, we want to speed-up calibration time
			 * so we privilege non-calibrated tasks (but still
			 * greedily distribute them to avoid dumb schedules) */
			calibrating = 1;

		if (local_task_length[worker] <= 0.0)
			/* there is no prediction available for that task
			 * with that arch yet, so switch to a greedy strategy */
			unknown = 1;

		if (unknown)
			continue;

		local_exp_end[worker] = compute_expected_end(worker, local_task_length[worker]);

		//fprintf(stderr, "WORKER %d -> length %e end %e\n", worker, local_task_length[worker], local_exp_end[worker]);

		if (local_exp_end[worker] < best_exp_end)
		{
			/* a better solution was found */
			best_exp_end = local_exp_end[worker];
		}

		local_power[worker] = starpu_task_expected_power(task, perf_arch);
		if (local_power[worker] == -1.0)
			local_power[worker] = 0.;
	}

	if (unknown)
		forced_best = ntasks_best;

	double best_fitness = -1;


	if (forced_best == -1)
	{
		for (worker = 0; worker < nworkers+ncombinedworkers; worker++)
		{

			if (skip_worker[worker])
			{
				/* no one on that queue may execute this task */
				continue;
			}
	
			fitness[worker] = alpha*(local_exp_end[worker] - best_exp_end) 
					+ beta*(local_data_penalty[worker])
					+ _gamma*(local_power[worker]);

			if (local_exp_end[worker] > max_exp_end)
				/* This placement will make the computation
				 * longer, take into account the idle
				 * consumption of other cpus */
				fitness[worker] += _gamma * idle_power * (local_exp_end[worker] - max_exp_end) / 1000000.0;

			if (best == -1 || fitness[worker] < best_fitness)
			{
				/* we found a better solution */
				best_fitness = fitness[worker];
				best = worker;
			}

		//	fprintf(stderr, "FITNESS worker %d -> %e local_exp_end %e - local_data_penalty %e\n", worker, fitness[worker], local_exp_end[worker] - best_exp_end, local_data_penalty[worker]);
		}
	}

	STARPU_ASSERT(forced_best != -1 || best != -1);

	if (forced_best != -1)
	{
		/* there is no prediction available for that task
		 * with that arch we want to speed-up calibration time
		 * so we force this measurement */
		best = forced_best;
		//penality_best = 0.0;
		best_exp_end = local_exp_end[best];
	}
	else 
	{
                //penality_best = local_data_penalty[best];
		best_exp_end = local_exp_end[best];
	}

	/* we should now have the best worker in variable "best" */
	return push_task_on_best_worker(task, best, best_exp_end, prio);
}

static int parallel_heft_push_prio_task(struct starpu_task *task)
{
	return _parallel_heft_push_task(task, 1);
}

static int parallel_heft_push_task(struct starpu_task *task)
{
	if (task->priority == STARPU_MAX_PRIO)
		return _parallel_heft_push_task(task, 1);

	return _parallel_heft_push_task(task, 0);
}

static void initialize_parallel_heft_policy(struct starpu_machine_topology_s *topology, 
	 __attribute__ ((unused)) struct starpu_sched_policy_s *_policy) 
{
	nworkers = topology->nworkers;

	const char *strval_alpha = getenv("STARPU_SCHED_ALPHA");
	if (strval_alpha)
		alpha = atof(strval_alpha);

	const char *strval_beta = getenv("STARPU_SCHED_BETA");
	if (strval_beta)
		beta = atof(strval_beta);

	const char *strval_gamma = getenv("STARPU_SCHED_GAMMA");
	if (strval_gamma)
		_gamma = atof(strval_gamma);

	const char *strval_idle_power = getenv("STARPU_IDLE_POWER");
	if (strval_idle_power)
		idle_power = atof(strval_idle_power);

	_starpu_sched_find_worker_combinations(topology);

	ncombinedworkers = topology->ncombinedworkers;

	unsigned workerid;
	for (workerid = 0; workerid < nworkers; workerid++)
	{
		worker_exp_start[workerid] = starpu_timing_now();
		worker_exp_len[workerid] = 0.0;
		worker_exp_end[workerid] = worker_exp_start[workerid]; 
		ntasks[workerid] = 0;
	
		PTHREAD_MUTEX_INIT(&sched_mutex[workerid], NULL);
		PTHREAD_COND_INIT(&sched_cond[workerid], NULL);
	
		starpu_worker_set_sched_condition(workerid, &sched_cond[workerid], &sched_mutex[workerid]);
	}

	PTHREAD_MUTEX_INIT(&big_lock, NULL);

	/* We pre-compute an array of all the perfmodel archs that are applicable */
	unsigned total_worker_count = nworkers + ncombinedworkers;

	unsigned used_perf_archtypes[STARPU_NARCH_VARIATIONS];
	memset(used_perf_archtypes, 0, sizeof(used_perf_archtypes));

	for (workerid = 0; workerid < total_worker_count; workerid++)
	{
		enum starpu_perf_archtype perf_archtype = starpu_worker_get_perf_archtype(workerid);
		used_perf_archtypes[perf_archtype] = 1;
	}

//	napplicable_perf_archtypes = 0;

//	int arch;
//	for (arch = 0; arch < STARPU_NARCH_VARIATIONS; arch++)
//	{
//		if (used_perf_archtypes[arch])
//			applicable_perf_archtypes[napplicable_perf_archtypes++] = arch;
//	}
}

/* TODO: use post_exec_hook to fix the expected start */
struct starpu_sched_policy_s _starpu_sched_parallel_heft_policy = {
	.init_sched = initialize_parallel_heft_policy,
	.deinit_sched = NULL,
	.push_task = parallel_heft_push_task, 
	.push_prio_task = parallel_heft_push_prio_task, 
	.pop_task = NULL,
	.post_exec_hook = parallel_heft_post_exec_hook,
	.pop_every_task = NULL,
	.policy_name = "pheft",
	.policy_description = "parallel HEFT"
};