backport branches/perfmodel@13175: src/drivers: use a more meaningful name than args for the worker parameter

src/drivers/cuda/driver_cuda.c

@@ -2,7 +2,7 @@
  *
  * Copyright (C) 2009-2014  Université de Bordeaux 1
  * Copyright (C) 2010  Mehdi Juhoor <mjuhoor@gmail.com>
- * Copyright (C) 2010, 2011, 2012, 2013  Centre National de la Recherche Scientifique
+ * Copyright (C) 2010, 2011, 2012, 2013, 2014  Centre National de la Recherche Scientifique
  * Copyright (C) 2011  Télécom-SudParis
  *
  * StarPU is free software; you can redistribute it and/or modify
@@ -372,7 +372,7 @@ void _starpu_init_cuda(void)
 	STARPU_ASSERT(ncudagpus <= STARPU_MAXCUDADEVS);
 }
 
-static int start_job_on_cuda(struct _starpu_job *j, struct _starpu_worker *args)
+static int start_job_on_cuda(struct _starpu_job *j, struct _starpu_worker *worker)
 {
 	int ret;
 
@@ -396,11 +396,11 @@ static int start_job_on_cuda(struct _starpu_job *j, struct _starpu_worker *args)
 		return -EAGAIN;
 	}
 
-	_starpu_driver_start_job(args, j, &args->perf_arch, &j->cl_start, 0, profiling);
+	_starpu_driver_start_job(worker, j, &worker->perf_arch, &j->cl_start, 0, profiling);
 
 #if defined(HAVE_CUDA_MEMCPY_PEER) && !defined(STARPU_SIMGRID)
 	/* We make sure we do manipulate the proper device */
-	starpu_cuda_set_device(args->devid);
+	starpu_cuda_set_device(worker->devid);
 #endif
 
 	starpu_cuda_func_t func = _starpu_task_get_cuda_nth_implementation(cl, j->nimpl);
@@ -410,7 +410,7 @@ static int start_job_on_cuda(struct _starpu_job *j, struct _starpu_worker *args)
 	{
 		_STARPU_TRACE_START_EXECUTING();
 #ifdef STARPU_SIMGRID
-		_starpu_simgrid_execute_job(j, &args->perf_arch, NAN);
+		_starpu_simgrid_execute_job(j, &worker->perf_arch, NAN);
 #else
 		func(_STARPU_TASK_GET_INTERFACES(task), task->cl_arg);
 #endif
@@ -420,18 +420,18 @@ static int start_job_on_cuda(struct _starpu_job *j, struct _starpu_worker *args)
 	return 0;
 }
 
-static void finish_job_on_cuda(struct _starpu_job *j, struct _starpu_worker *args)
+static void finish_job_on_cuda(struct _starpu_job *j, struct _starpu_worker *worker)
 {
 	struct timespec codelet_end;
 
 	int profiling = starpu_profiling_status_get();
 
 	_starpu_set_current_task(NULL);
-	args->current_task = NULL;
+	worker->current_task = NULL;
 
-	_starpu_driver_end_job(args, j, &args->perf_arch, &codelet_end, 0, profiling);
+	_starpu_driver_end_job(worker, j, &worker->perf_arch, &codelet_end, 0, profiling);
 
-	_starpu_driver_update_job_feedback(j, args, &args->perf_arch, &j->cl_start, &codelet_end, profiling);
+	_starpu_driver_update_job_feedback(j, worker, &worker->perf_arch, &j->cl_start, &codelet_end, profiling);
 
 	_starpu_push_task_output(j);
 
@@ -441,18 +441,18 @@ static void finish_job_on_cuda(struct _starpu_job *j, struct _starpu_worker *arg
 /* XXX Should this be merged with _starpu_init_cuda ? */
 int _starpu_cuda_driver_init(struct _starpu_worker_set *worker_set)
 {
-	struct _starpu_worker *args = &worker_set->workers[0];
-	unsigned devid = args->devid;
+	struct _starpu_worker *worker = &worker_set->workers[0];
+	unsigned devid = worker->devid;
 	unsigned i;
 
-	_starpu_worker_start(args, _STARPU_FUT_CUDA_KEY);
+	_starpu_worker_start(worker, _STARPU_FUT_CUDA_KEY);
 
 #ifdef STARPU_USE_FXT
-	unsigned memnode = args->memory_node;
+	unsigned memnode = worker->memory_node;
 	for (i = 1; i < worker_set->nworkers; i++)
 	{
-		struct _starpu_worker *worker = &worker_set->workers[i];
-		_STARPU_TRACE_WORKER_INIT_START(_STARPU_FUT_CUDA_KEY, worker->workerid, devid, memnode);
+		struct _starpu_worker *_worker = &worker_set->workers[i];
+		_STARPU_TRACE_WORKER_INIT_START(_STARPU_FUT_CUDA_KEY, _worker->workerid, devid, memnode);
 	}
 #endif
 
@@ -461,14 +461,14 @@ int _starpu_cuda_driver_init(struct _starpu_worker_set *worker_set)
 #endif
 
 	_starpu_cuda_limit_gpu_mem_if_needed(devid);
-	_starpu_memory_manager_set_global_memory_size(args->memory_node, _starpu_cuda_get_global_mem_size(devid));
+	_starpu_memory_manager_set_global_memory_size(worker->memory_node, _starpu_cuda_get_global_mem_size(devid));
 
-	_starpu_malloc_init(args->memory_node);
+	_starpu_malloc_init(worker->memory_node);
 
 	/* one more time to avoid hacks from third party lib :) */
-	_starpu_bind_thread_on_cpu(args->config, args->bindid);
+	_starpu_bind_thread_on_cpu(worker->config, worker->bindid);
 
-	args->status = STATUS_UNKNOWN;
+	worker->status = STATUS_UNKNOWN;
 
 	float size = (float) global_mem[devid] / (1<<30);
 #ifdef STARPU_SIMGRID
@@ -482,15 +482,15 @@ int _starpu_cuda_driver_init(struct _starpu_worker_set *worker_set)
 #if defined(STARPU_HAVE_BUSID) && !defined(STARPU_SIMGRID)
 #if defined(STARPU_HAVE_DOMAINID) && !defined(STARPU_SIMGRID)
 	if (props[devid].pciDomainID)
-		snprintf(args->name, sizeof(args->name), "CUDA %u (%s %.1f GiB %04x:%02x:%02x.0)", devid, devname, size, props[devid].pciDomainID, props[devid].pciBusID, props[devid].pciDeviceID);
+		snprintf(worker->name, sizeof(worker->name), "CUDA %u (%s %.1f GiB %04x:%02x:%02x.0)", devid, devname, size, props[devid].pciDomainID, props[devid].pciBusID, props[devid].pciDeviceID);
 	else
 #endif
-		snprintf(args->name, sizeof(args->name), "CUDA %u (%s %.1f GiB %02x:%02x.0)", devid, devname, size, props[devid].pciBusID, props[devid].pciDeviceID);
+		snprintf(worker->name, sizeof(worker->name), "CUDA %u (%s %.1f GiB %02x:%02x.0)", devid, devname, size, props[devid].pciBusID, props[devid].pciDeviceID);
 #else
-	snprintf(args->name, sizeof(args->name), "CUDA %u (%s %.1f GiB)", devid, devname, size);
+	snprintf(worker->name, sizeof(worker->name), "CUDA %u (%s %.1f GiB)", devid, devname, size);
 #endif
-	snprintf(args->short_name, sizeof(args->short_name), "CUDA %u", devid);
-	_STARPU_DEBUG("cuda (%s) dev id %u thread is ready to run on CPU %d !\n", devname, devid, args->bindid);
+	snprintf(worker->short_name, sizeof(worker->short_name), "CUDA %u", devid);
+	_STARPU_DEBUG("cuda (%s) dev id %u thread is ready to run on CPU %d !\n", devname, devid, worker->bindid);
 
 	for (i = 0; i < worker_set->nworkers; i++)
 	{
@@ -498,10 +498,10 @@ int _starpu_cuda_driver_init(struct _starpu_worker_set *worker_set)
 	}
 
 	/* tell the main thread that this one is ready */
-	STARPU_PTHREAD_MUTEX_LOCK(&args->mutex);
-	args->worker_is_initialized = 1;
-	STARPU_PTHREAD_COND_SIGNAL(&args->ready_cond);
-	STARPU_PTHREAD_MUTEX_UNLOCK(&args->mutex);
+	STARPU_PTHREAD_MUTEX_LOCK(&worker->mutex);
+	worker->worker_is_initialized = 1;
+	STARPU_PTHREAD_COND_SIGNAL(&worker->ready_cond);
+	STARPU_PTHREAD_MUTEX_UNLOCK(&worker->mutex);
 
 	/* tell the main thread that this one is ready */
 	STARPU_PTHREAD_MUTEX_LOCK(&worker_set->mutex);
@@ -527,10 +527,10 @@ int _starpu_cuda_driver_run_once(struct _starpu_worker_set *worker_set)
 	idle = 0;
 	for (i = 0; i < (int) worker_set->nworkers; i++)
 	{
-		struct _starpu_worker *args = &worker_set->workers[i];
-		int workerid = args->workerid;
+		struct _starpu_worker *worker = &worker_set->workers[i];
+		int workerid = worker->workerid;
 
-		task = args->current_task;
+		task = worker->current_task;
 
 		if (!task)
 		{
@@ -548,8 +548,8 @@ int _starpu_cuda_driver_run_once(struct _starpu_worker_set *worker_set)
 		else
 		{
 			/* Asynchronous task completed! */
-			_starpu_set_local_worker_key(args);
-			finish_job_on_cuda(_starpu_get_job_associated_to_task(task), args);
+			_starpu_set_local_worker_key(worker);
+			finish_job_on_cuda(_starpu_get_job_associated_to_task(task), worker);
 			idle++;
 #ifdef STARPU_USE_FXT
 			int k;
@@ -582,14 +582,14 @@ int _starpu_cuda_driver_run_once(struct _starpu_worker_set *worker_set)
 
 	for (i = 0; i < (int) worker_set->nworkers; i++)
 	{
-		struct _starpu_worker *args = &worker_set->workers[i];
-		int workerid = args->workerid;
+		struct _starpu_worker *worker = &worker_set->workers[i];
+		int workerid = worker->workerid;
 
 		task = tasks[i];
 		if (!task)
 			continue;
 
-		_starpu_set_local_worker_key(args);
+		_starpu_set_local_worker_key(worker);
 
 		j = _starpu_get_job_associated_to_task(task);
 
@@ -602,7 +602,7 @@ int _starpu_cuda_driver_run_once(struct _starpu_worker_set *worker_set)
 		}
 
 		_STARPU_TRACE_END_PROGRESS(memnode);
-		res = start_job_on_cuda(j, args);
+		res = start_job_on_cuda(j, worker);
 
 		if (res)
 		{
@@ -643,7 +643,7 @@ int _starpu_cuda_driver_run_once(struct _starpu_worker_set *worker_set)
 #if defined(STARPU_DEBUG) && !defined(STARPU_SIMGRID)
 			STARPU_ASSERT_MSG(cudaStreamQuery(starpu_cuda_get_local_stream()) == cudaSuccess, "CUDA codelets have to wait for termination of their kernels on the starpu_cuda_get_local_stream() stream");
 #endif
-			finish_job_on_cuda(j, args);
+			finish_job_on_cuda(j, worker);
 		}
 		_STARPU_TRACE_START_PROGRESS(memnode);
 	}
@@ -653,8 +653,8 @@ int _starpu_cuda_driver_run_once(struct _starpu_worker_set *worker_set)
 
 int _starpu_cuda_driver_deinit(struct _starpu_worker_set *arg)
 {
-	struct _starpu_worker *args = &arg->workers[0];
-	unsigned memnode = args->memory_node;
+	struct _starpu_worker *worker = &arg->workers[0];
+	unsigned memnode = worker->memory_node;
 	_STARPU_TRACE_WORKER_DEINIT_START;
 
 	_starpu_handle_all_pending_node_data_requests(memnode);
@@ -675,16 +675,16 @@ int _starpu_cuda_driver_deinit(struct _starpu_worker_set *arg)
 	return 0;
 }
 
-void *_starpu_cuda_worker(void *arg)
+void *_starpu_cuda_worker(void *_arg)
 {
-	struct _starpu_worker_set* args = arg;
+	struct _starpu_worker_set* worker = _arg;
 
-	_starpu_cuda_driver_init(args);
+	_starpu_cuda_driver_init(worker);
 	_STARPU_TRACE_START_PROGRESS(memnode);
 	while (_starpu_machine_is_running())
-		_starpu_cuda_driver_run_once(args);
+		_starpu_cuda_driver_run_once(worker);
 	_STARPU_TRACE_END_PROGRESS(memnode);
-	_starpu_cuda_driver_deinit(args);
+	_starpu_cuda_driver_deinit(worker);
 
 	return NULL;
 }

src/drivers/driver_common/driver_common.c

@@ -1,7 +1,7 @@
 /* StarPU --- Runtime system for heterogeneous multicore architectures.
  *
  * Copyright (C) 2010-2014  Université de Bordeaux 1
- * Copyright (C) 2010, 2011, 2012, 2013  Centre National de la Recherche Scientifique
+ * Copyright (C) 2010, 2011, 2012, 2013, 2014  Centre National de la Recherche Scientifique
  * Copyright (C) 2011  Télécom-SudParis
  *
  * StarPU is free software; you can redistribute it and/or modify
@@ -33,13 +33,13 @@
 #define BACKOFF_MAX 32  /* TODO : use parameter to define them */
 #define BACKOFF_MIN 1
 
-void _starpu_driver_start_job(struct _starpu_worker *args, struct _starpu_job *j, struct starpu_perfmodel_arch* perf_arch STARPU_ATTRIBUTE_UNUSED, struct timespec *codelet_start, int rank, int profiling)
+void _starpu_driver_start_job(struct _starpu_worker *worker, struct _starpu_job *j, struct starpu_perfmodel_arch* perf_arch STARPU_ATTRIBUTE_UNUSED, struct timespec *codelet_start, int rank, int profiling)
 {
 	struct starpu_task *task = j->task;
 	struct starpu_codelet *cl = task->cl;
 	struct starpu_profiling_task_info *profiling_info;
 	int starpu_top=_starpu_top_status_get();
-	int workerid = args->workerid;
+	int workerid = worker->workerid;
 	unsigned calibrate_model = 0;
 
 	if (cl->model && cl->model->benchmarking)
@@ -51,7 +51,7 @@ void _starpu_driver_start_job(struct _starpu_worker *args, struct _starpu_job *j
 	if (j->task_size == 1)
 		_starpu_sched_pre_exec_hook(task);
 
-	args->status = STATUS_EXECUTING;
+	worker->status = STATUS_EXECUTING;
 	task->status = STARPU_TASK_RUNNING;
 
 	if (rank == 0)
@@ -76,13 +76,13 @@ void _starpu_driver_start_job(struct _starpu_worker *args, struct _starpu_job *j
 	_STARPU_TRACE_START_CODELET_BODY(j, j->nimpl, perf_arch, workerid);
 }
 
-void _starpu_driver_end_job(struct _starpu_worker *args, struct _starpu_job *j, struct starpu_perfmodel_arch* perf_arch STARPU_ATTRIBUTE_UNUSED, struct timespec *codelet_end, int rank, int profiling)
+void _starpu_driver_end_job(struct _starpu_worker *worker, struct _starpu_job *j, struct starpu_perfmodel_arch* perf_arch STARPU_ATTRIBUTE_UNUSED, struct timespec *codelet_end, int rank, int profiling)
 {
 	struct starpu_task *task = j->task;
 	struct starpu_codelet *cl = task->cl;
 	struct starpu_profiling_task_info *profiling_info = task->profiling_info;
 	int starpu_top=_starpu_top_status_get();
-	int workerid = args->workerid;
+	int workerid = worker->workerid;
 	unsigned calibrate_model = 0;
 
 	_STARPU_TRACE_END_CODELET_BODY(j, j->nimpl, perf_arch, workerid);
@@ -102,16 +102,16 @@ void _starpu_driver_end_job(struct _starpu_worker *args, struct _starpu_job *j,
 	if (starpu_top)
 		_starpu_top_task_ended(task,workerid,codelet_end);
 
-	args->status = STATUS_UNKNOWN;
+	worker->status = STATUS_UNKNOWN;
 }
-void _starpu_driver_update_job_feedback(struct _starpu_job *j, struct _starpu_worker *worker_args,
+void _starpu_driver_update_job_feedback(struct _starpu_job *j, struct _starpu_worker *worker,
 					struct starpu_perfmodel_arch* perf_arch,
 					struct timespec *codelet_start, struct timespec *codelet_end, int profiling)
 {
 	struct starpu_profiling_task_info *profiling_info = j->task->profiling_info;
 	struct timespec measured_ts;
 	double measured;
-	int workerid = worker_args->workerid;
+	int workerid = worker->workerid;
 	struct starpu_codelet *cl = j->task->cl;
 	int calibrate_model = 0;
 	int updated = 0;
@@ -141,7 +141,7 @@ void _starpu_driver_update_job_feedback(struct _starpu_job *j, struct _starpu_wo
 		}
 
 		if (calibrate_model)
-			_starpu_update_perfmodel_history(j, j->task->cl->model,  perf_arch, worker_args->devid, measured,j->nimpl);
+			_starpu_update_perfmodel_history(j, j->task->cl->model,  perf_arch, worker->devid, measured,j->nimpl);
 
 
 	}
@@ -151,7 +151,7 @@ void _starpu_driver_update_job_feedback(struct _starpu_job *j, struct _starpu_wo
 
 	if (profiling_info && profiling_info->power_consumed && cl->power_model && cl->power_model->benchmarking)
 	{
-		_starpu_update_perfmodel_history(j, j->task->cl->power_model, perf_arch, worker_args->devid, profiling_info->power_consumed,j->nimpl);
+		_starpu_update_perfmodel_history(j, j->task->cl->power_model, perf_arch, worker->devid, profiling_info->power_consumed,j->nimpl);
 	}
 }
 
@@ -198,12 +198,12 @@ static void _starpu_worker_set_status_wakeup(int workerid)
 }
 
 
-static void _starpu_exponential_backoff(struct _starpu_worker *args)
+static void _starpu_exponential_backoff(struct _starpu_worker *worker)
 {
-	int delay = args->spinning_backoff;
+	int delay = worker->spinning_backoff;
 	
-	if (args->spinning_backoff < BACKOFF_MAX)
-		args->spinning_backoff<<=1; 
+	if (worker->spinning_backoff < BACKOFF_MAX)
+		worker->spinning_backoff<<=1; 
 	
 	while(delay--)
 		STARPU_UYIELD();
@@ -212,9 +212,9 @@ static void _starpu_exponential_backoff(struct _starpu_worker *args)
 
 
 /* Workers may block when there is no work to do at all. */
-struct starpu_task *_starpu_get_worker_task(struct _starpu_worker *args, int workerid, unsigned memnode)
+struct starpu_task *_starpu_get_worker_task(struct _starpu_worker *worker, int workerid, unsigned memnode)
 {
-	STARPU_PTHREAD_MUTEX_LOCK(&args->sched_mutex);
+	STARPU_PTHREAD_MUTEX_LOCK(&worker->sched_mutex);
 	struct starpu_task *task;
 	unsigned needed = 1;
 	_starpu_worker_set_status_scheduling(workerid);
@@ -222,7 +222,7 @@ struct starpu_task *_starpu_get_worker_task(struct _starpu_worker *args, int wor
 	{
 		struct _starpu_sched_ctx *sched_ctx = NULL;
 		struct _starpu_sched_ctx_list *l = NULL;
-		for (l = args->sched_ctx_list; l; l = l->next)
+		for (l = worker->sched_ctx_list; l; l = l->next)
 		{
 			sched_ctx = _starpu_get_sched_ctx_struct(l->sched_ctx);
 			if(sched_ctx && sched_ctx->id > 0 && sched_ctx->id < STARPU_NMAX_SCHED_CTXS)
@@ -233,13 +233,13 @@ struct starpu_task *_starpu_get_worker_task(struct _starpu_worker *args, int wor
 					/* don't let the worker sleep with the sched_mutex taken */
 					/* we need it until here bc of the list of ctxs of the workers
 					   that can change in another thread */
-					STARPU_PTHREAD_MUTEX_UNLOCK(&args->sched_mutex);
+					STARPU_PTHREAD_MUTEX_UNLOCK(&worker->sched_mutex);
 					needed = 0;
 					_starpu_sched_ctx_signal_worker_blocked(sched_ctx->id, workerid);
 					STARPU_PTHREAD_COND_WAIT(&sched_ctx->parallel_sect_cond[workerid], &sched_ctx->parallel_sect_mutex[workerid]);
 					_starpu_sched_ctx_signal_worker_woke_up(sched_ctx->id, workerid);
 					sched_ctx->parallel_sect[workerid] = 0;
-					STARPU_PTHREAD_MUTEX_LOCK(&args->sched_mutex);
+					STARPU_PTHREAD_MUTEX_LOCK(&worker->sched_mutex);
 				}
 				STARPU_PTHREAD_MUTEX_UNLOCK(&sched_ctx->parallel_sect_mutex[workerid]);
 			}
@@ -247,19 +247,19 @@ struct starpu_task *_starpu_get_worker_task(struct _starpu_worker *args, int wor
 				break;
 		}
 		/* don't worry if the value is not correct (no lock) it will do it next time */
-		if(args->tmp_sched_ctx != -1)
+		if(worker->tmp_sched_ctx != -1)
 		{
-			sched_ctx = _starpu_get_sched_ctx_struct(args->tmp_sched_ctx);
+			sched_ctx = _starpu_get_sched_ctx_struct(worker->tmp_sched_ctx);
 			STARPU_PTHREAD_MUTEX_LOCK(&sched_ctx->parallel_sect_mutex[workerid]);
 			if(sched_ctx->parallel_sect[workerid])
 			{
 //				needed = 0;
-				STARPU_PTHREAD_MUTEX_UNLOCK(&args->sched_mutex);
+				STARPU_PTHREAD_MUTEX_UNLOCK(&worker->sched_mutex);
 				_starpu_sched_ctx_signal_worker_blocked(sched_ctx->id, workerid);
 				STARPU_PTHREAD_COND_WAIT(&sched_ctx->parallel_sect_cond[workerid], &sched_ctx->parallel_sect_mutex[workerid]);
 				_starpu_sched_ctx_signal_worker_woke_up(sched_ctx->id, workerid);
 				sched_ctx->parallel_sect[workerid] = 0;
-				STARPU_PTHREAD_MUTEX_LOCK(&args->sched_mutex);
+				STARPU_PTHREAD_MUTEX_LOCK(&worker->sched_mutex);
 			}
 			STARPU_PTHREAD_MUTEX_UNLOCK(&sched_ctx->parallel_sect_mutex[workerid]);
 		}
@@ -267,7 +267,7 @@ struct starpu_task *_starpu_get_worker_task(struct _starpu_worker *args, int wor
 		needed = !needed;
 	}
 
-	task = _starpu_pop_task(args);
+	task = _starpu_pop_task(worker);
 
 	if (task == NULL)
 	{
@@ -278,17 +278,17 @@ struct starpu_task *_starpu_get_worker_task(struct _starpu_worker *args, int wor
 
 		_starpu_worker_set_status_sleeping(workerid);
 
-		if (_starpu_worker_can_block(memnode) && !_starpu_sched_ctx_last_worker_awake(args))
+		if (_starpu_worker_can_block(memnode) && !_starpu_sched_ctx_last_worker_awake(worker))
 		{
-			STARPU_PTHREAD_COND_WAIT(&args->sched_cond, &args->sched_mutex);
-			STARPU_PTHREAD_MUTEX_UNLOCK(&args->sched_mutex);
+			STARPU_PTHREAD_COND_WAIT(&worker->sched_cond, &worker->sched_mutex);
+			STARPU_PTHREAD_MUTEX_UNLOCK(&worker->sched_mutex);
 		}
 		else
 		{
-			STARPU_PTHREAD_MUTEX_UNLOCK(&args->sched_mutex);			
+			STARPU_PTHREAD_MUTEX_UNLOCK(&worker->sched_mutex);
 			if (_starpu_machine_is_running())
 			{
-				_starpu_exponential_backoff(args);
+				_starpu_exponential_backoff(worker);
 #ifdef STARPU_SIMGRID
 				static int warned;
 				if (!warned)
@@ -307,9 +307,9 @@ struct starpu_task *_starpu_get_worker_task(struct _starpu_worker *args, int wor
 	_starpu_worker_set_status_scheduling_done(workerid);
 
 	_starpu_worker_set_status_wakeup(workerid);
-	args->spinning_backoff = BACKOFF_MIN;
+	worker->spinning_backoff = BACKOFF_MIN;
 
-	STARPU_PTHREAD_MUTEX_UNLOCK(&args->sched_mutex);
+	STARPU_PTHREAD_MUTEX_UNLOCK(&worker->sched_mutex);
 
 
 #ifdef HAVE_AYUDAME_H

src/drivers/opencl/driver_opencl.c

@@ -2,7 +2,7 @@
  *
  * Copyright (C) 2010-2014  Université de Bordeaux 1
  * Copyright (C) 2010  Mehdi Juhoor <mjuhoor@gmail.com>
- * Copyright (C) 2010, 2011, 2012, 2013  Centre National de la Recherche Scientifique
+ * Copyright (C) 2010, 2011, 2012, 2013, 2014  Centre National de la Recherche Scientifique
  * Copyright (C) 2011  Télécom-SudParis
  *
  * StarPU is free software; you can redistribute it and/or modify
@@ -562,8 +562,8 @@ void _starpu_opencl_init(void)
 #ifndef STARPU_SIMGRID
 static unsigned _starpu_opencl_get_device_name(int dev, char *name, int lname);
 #endif
-static int _starpu_opencl_start_job(struct _starpu_job *j, struct _starpu_worker *args);
-static void _starpu_opencl_stop_job(struct _starpu_job *j, struct _starpu_worker *args);
+static int _starpu_opencl_start_job(struct _starpu_job *j, struct _starpu_worker *worker);
+static void _starpu_opencl_stop_job(struct _starpu_job *j, struct _starpu_worker *worker);
 
 int _starpu_opencl_driver_init(struct _starpu_worker *args)
 {
@@ -609,10 +609,10 @@ int _starpu_opencl_driver_init(struct _starpu_worker *args)
 	return 0;
 }
 
-int _starpu_opencl_driver_run_once(struct _starpu_worker *args)
+int _starpu_opencl_driver_run_once(struct _starpu_worker *worker)
 {
-	int workerid = args->workerid;
-	unsigned memnode = args->memory_node;
+	int workerid = worker->workerid;
+	unsigned memnode = worker->memory_node;
 
 	struct _starpu_job *j;
 	struct starpu_task *task;
@@ -628,7 +628,7 @@ int _starpu_opencl_driver_run_once(struct _starpu_worker *args)
 		int err;
 		/* On-going asynchronous task, check for its termination first */
 
-		err = clGetEventInfo(task_events[args->devid], CL_EVENT_COMMAND_EXECUTION_STATUS, sizeof(cl_int), &status, &size);
+		err = clGetEventInfo(task_events[worker->devid], CL_EVENT_COMMAND_EXECUTION_STATUS, sizeof(cl_int), &status, &size);
 		STARPU_ASSERT(size == sizeof(cl_int));
 		if (STARPU_UNLIKELY(err != CL_SUCCESS)) STARPU_OPENCL_REPORT_ERROR(err);
 
@@ -642,7 +642,7 @@ int _starpu_opencl_driver_run_once(struct _starpu_worker *args)
 
 		/* Asynchronous task completed! */
 		_STARPU_TRACE_END_EXECUTING();
-		_starpu_opencl_stop_job(_starpu_get_job_associated_to_task(task), args);
+		_starpu_opencl_stop_job(_starpu_get_job_associated_to_task(task), worker);
 	}
 #endif /* STARPU_SIMGRID */
 
@@ -650,7 +650,7 @@ int _starpu_opencl_driver_run_once(struct _starpu_worker *args)
 
 	_STARPU_TRACE_END_PROGRESS(memnode);
 
-	task = _starpu_get_worker_task(args, workerid, memnode);
+	task = _starpu_get_worker_task(worker, workerid, memnode);
 
 	if (task == NULL)
 		return 0;
@@ -665,7 +665,7 @@ int _starpu_opencl_driver_run_once(struct _starpu_worker *args)
 		return 0;
 	}
 
-	res = _starpu_opencl_start_job(j, args);
+	res = _starpu_opencl_start_job(j, worker);
 
 	if (res)
 	{
@@ -687,7 +687,7 @@ int _starpu_opencl_driver_run_once(struct _starpu_worker *args)
 		/* Record event to synchronize with task termination later */
 		int err;
 		cl_command_queue queue;
-		starpu_opencl_get_queue(args->devid, &queue);
+		starpu_opencl_get_queue(worker->devid, &queue);
 		/* the function clEnqueueMarker is deprecated from
 		 * OpenCL version 1.2. We would like to use the new
 		 * function clEnqueueMarkerWithWaitList. We could do
@@ -697,7 +697,7 @@ int _starpu_opencl_driver_run_once(struct _starpu_worker *args)
 		 * 2 macros detect the function availability in the
 		 * ICD and not in the device implementation.
 		 */
-		err = clEnqueueMarker(queue, &task_events[args->devid]);
+		err = clEnqueueMarker(queue, &task_events[worker->devid]);
 		if (STARPU_UNLIKELY(err != CL_SUCCESS)) STARPU_OPENCL_REPORT_ERROR(err);
 		_STARPU_TRACE_START_EXECUTING();
 	}
@@ -709,18 +709,18 @@ int _starpu_opencl_driver_run_once(struct _starpu_worker *args)
 #endif
 	/* Synchronous execution */
 	{
-		_starpu_opencl_stop_job(j, args);
+		_starpu_opencl_stop_job(j, worker);
 	}
 	_STARPU_TRACE_START_PROGRESS(memnode);
 
 	return 0;
 }
 
-int _starpu_opencl_driver_deinit(struct _starpu_worker *args)
+int _starpu_opencl_driver_deinit(struct _starpu_worker *worker)
 {
 	_STARPU_TRACE_WORKER_DEINIT_START;
 
-	unsigned memnode = args->memory_node;
+	unsigned memnode = worker->memory_node;
 
 	_starpu_handle_all_pending_node_data_requests(memnode);
 
@@ -732,7 +732,7 @@ int _starpu_opencl_driver_deinit(struct _starpu_worker *args)
 	_starpu_malloc_shutdown(memnode);
 
 #ifndef STARPU_SIMGRID
-	unsigned devid   = args->devid;
+	unsigned devid   = worker->devid;
         _starpu_opencl_deinit_context(devid);
 #endif
 
@@ -741,15 +741,15 @@ int _starpu_opencl_driver_deinit(struct _starpu_worker *args)
 	return 0;
 }
 
-void *_starpu_opencl_worker(void *arg)
+void *_starpu_opencl_worker(void *_arg)
 {
-	struct _starpu_worker* args = arg;
+	struct _starpu_worker* worker = _arg;
 
-	_starpu_opencl_driver_init(args);
+	_starpu_opencl_driver_init(worker);
 	_STARPU_TRACE_START_PROGRESS(memnode);
 	while (_starpu_machine_is_running())
-		_starpu_opencl_driver_run_once(args);
-	_starpu_opencl_driver_deinit(args);
+		_starpu_opencl_driver_run_once(worker);
+	_starpu_opencl_driver_deinit(worker);
 	_STARPU_TRACE_END_PROGRESS(memnode);
 
 	return NULL;
@@ -802,7 +802,7 @@ cl_device_type _starpu_opencl_get_device_type(int devid)
 }
 #endif /* STARPU_USE_OPENCL */
 
-static int _starpu_opencl_start_job(struct _starpu_job *j, struct _starpu_worker *args)
+static int _starpu_opencl_start_job(struct _starpu_job *j, struct _starpu_worker *worker)
 {
 	int ret;
 
@@ -816,7 +816,7 @@ static int _starpu_opencl_start_job(struct _starpu_job *j, struct _starpu_worker
 	STARPU_ASSERT(cl);
 
 	_starpu_set_current_task(j->task);
-	args->current_task = j->task;
+	worker->current_task = j->task;
 
 	ret = _starpu_fetch_task_input(j);
 	if (ret != 0)
@@ -827,7 +827,7 @@ static int _starpu_opencl_start_job(struct _starpu_job *j, struct _starpu_worker
 		return -EAGAIN;
 	}
 
-	_starpu_driver_start_job(args, j, &args->perf_arch, &j->cl_start, 0, profiling);
+	_starpu_driver_start_job(worker, j, &worker->perf_arch, &j->cl_start, 0, profiling);
 
 	starpu_opencl_func_t func = _starpu_task_get_opencl_nth_implementation(cl, j->nimpl);
 	STARPU_ASSERT_MSG(func, "when STARPU_OPENCL is defined in 'where', opencl_func or opencl_funcs has to be defined");
@@ -850,7 +850,7 @@ static int _starpu_opencl_start_job(struct _starpu_job *j, struct _starpu_worker
 		STARPU_ASSERT_MSG(profiling_info->used_cycles, "Application kernel must call starpu_opencl_collect_stats to collect simulated time");
 		length = ((double) profiling_info->used_cycles)/MSG_get_host_speed(MSG_host_self());
 	  #endif
-		_starpu_simgrid_execute_job(j, &args->perf_arch, length);
+		_starpu_simgrid_execute_job(j, &worker->perf_arch, length);
 #else
 		func(_STARPU_TASK_GET_INTERFACES(task), task->cl_arg);
 #endif
@@ -859,18 +859,17 @@ static int _starpu_opencl_start_job(struct _starpu_job *j, struct _starpu_worker
 	return 0;
 }
 
-static void _starpu_opencl_stop_job(struct _starpu_job *j, struct _starpu_worker *args)
+static void _starpu_opencl_stop_job(struct _starpu_job *j, struct _starpu_worker *worker)
 {
 	struct timespec codelet_end;
 	int profiling = starpu_profiling_status_get();
 
 	_starpu_set_current_task(NULL);
-	args->current_task = NULL;
+	worker->current_task = NULL;
 
-	_starpu_driver_end_job(args, j, &args->perf_arch, &codelet_end, 0, profiling);
+	_starpu_driver_end_job(worker, j, &worker->perf_arch, &codelet_end, 0, profiling);
 
-	_starpu_driver_update_job_feedback(j, args, &args->perf_arch,
-					   &j->cl_start, &codelet_end, profiling);
+	_starpu_driver_update_job_feedback(j, worker, &worker->perf_arch, &j->cl_start, &codelet_end, profiling);
 
 	_starpu_push_task_output(j);