merge trunk

configure.ac

@@ -88,8 +88,18 @@ AC_C_RESTRICT
 AC_CHECK_PROGS([BASH], [bash])
 
 # Check whether subversion is installed
+AC_PATH_PROG(svncommand, svn)
 AC_PATH_PROG(svnversioncommand, svnversion)
 
+# find out if we are are in a subversion directory
+svndir=0
+if test "$svncommand" != "" ; then
+   $svncommand info $srcdir >/dev/null 2>&1
+   if test $? -eq 0; then
+      svndir=1
+   fi
+fi
+
 # use svnversion to record the current repository revision only if
 # subversion is installed and we are in a working copy
 if test "$svnversioncommand" = "" || test "`LC_ALL=C $svnversioncommand -n $srcdir`" = "exported" ; then
@@ -1761,7 +1771,7 @@ AC_SUBST(CC_OR_MPICC, $cc_or_mpicc)
 # If the user specifically asks for it, or if we are in a developer checkout, we enable mpi check
 AC_ARG_ENABLE(mpi-check, AC_HELP_STRING([--enable-mpi-check], [Enable execution of MPI testcases]))
 running_mpi_check=no
-if test -d "$srcdir/.svn" -o -d "$srcdir/.git" ; then
+if test $svndir = 1 -o -d "$srcdir/.git" ; then
     running_mpi_check=yes
 fi
 if test x$enable_mpi_check = xyes ; then

doc/doxygen/chapters/08scheduling.doxy

@@ -128,6 +128,10 @@ the user can for instance run a given task a thousand times, measure the global
 consumption for that series of tasks, divide it by a thousand, repeat for
 varying kinds of tasks and task sizes, and eventually feed StarPU
 with these manual measurements through starpu_perfmodel_update_history().
+For instance, for CUDA devices, <c>nvidia-smi -q -d POWER</c> can be used to get
+the current consumption in Watt. Multiplying that value by the average duration
+of a single task gives the consumption of the task in Joules, which can be given
+to starpu_perfmodel_update_history().
 
 \section StaticScheduling Static Scheduling
 

src/core/sched_ctx.c

@@ -23,13 +23,13 @@ starpu_pthread_rwlock_t changing_ctx_mutex[STARPU_NMAX_SCHED_CTXS];
 
 static starpu_pthread_mutex_t sched_ctx_manag = STARPU_PTHREAD_MUTEX_INITIALIZER;
 static starpu_pthread_mutex_t finished_submit_mutex = STARPU_PTHREAD_MUTEX_INITIALIZER;
-struct starpu_task stop_submission_task = STARPU_TASK_INITIALIZER;
+static struct starpu_task stop_submission_task = STARPU_TASK_INITIALIZER;
 starpu_pthread_key_t sched_ctx_key;
-unsigned with_hypervisor = 0;
-double hyp_start_sample[STARPU_NMAX_SCHED_CTXS];
-double hyp_start_allow_sample[STARPU_NMAX_SCHED_CTXS];
-double flops[STARPU_NMAX_SCHED_CTXS][STARPU_NMAXWORKERS];
-size_t data_size[STARPU_NMAX_SCHED_CTXS][STARPU_NMAXWORKERS];
+static unsigned with_hypervisor = 0;
+static double hyp_start_sample[STARPU_NMAX_SCHED_CTXS];
+static double hyp_start_allow_sample[STARPU_NMAX_SCHED_CTXS];
+static double flops[STARPU_NMAX_SCHED_CTXS][STARPU_NMAXWORKERS];
+static size_t data_size[STARPU_NMAX_SCHED_CTXS][STARPU_NMAXWORKERS];
 
 static unsigned _starpu_get_first_free_sched_ctx(struct _starpu_machine_config *config);
 static void _starpu_sched_ctx_add_workers_to_master(unsigned sched_ctx_id, int *workerids, int nworkers, int new_master);
@@ -49,7 +49,7 @@ static void _starpu_worker_gets_into_ctx(unsigned sched_ctx_id, struct _starpu_w
 		worker->nsched_ctxs++;
 	}
 	worker->removed_from_ctx[sched_ctx_id] = 0;
-	if(worker->tmp_sched_ctx == sched_ctx_id)
+	if(worker->tmp_sched_ctx == (int) sched_ctx_id)
 		worker->tmp_sched_ctx = -1;
 	return;
 }
@@ -1842,10 +1842,10 @@ struct _starpu_sched_ctx *_starpu_sched_ctx_get_sched_ctx_for_worker_and_job(str
 	return NULL;
 }
 
-void starpu_sched_ctx_revert_task_counters(unsigned sched_ctx_id, double flops)
+void starpu_sched_ctx_revert_task_counters(unsigned sched_ctx_id, double ready_flops)
 {
         _starpu_decrement_nsubmitted_tasks_of_sched_ctx(sched_ctx_id);
-        _starpu_decrement_nready_tasks_of_sched_ctx(sched_ctx_id, flops);
+        _starpu_decrement_nready_tasks_of_sched_ctx(sched_ctx_id, ready_flops);
 }
 
 void starpu_sched_ctx_move_task_to_ctx(struct starpu_task *task, unsigned sched_ctx)
@@ -1989,7 +1989,6 @@ void starpu_sched_ctx_get_available_cpuids(unsigned sched_ctx_id, int **cpuids,
 	(*cpuids) = (int*)malloc(workers->nworkers*sizeof(int));
 	int w = 0;
 
-	struct _starpu_worker *worker = NULL;
 	struct starpu_sched_ctx_iterator it;
 	int workerid;
 	if(workers->init_iterator)
@@ -2015,7 +2014,6 @@ static void _starpu_sched_ctx_wake_these_workers_up(unsigned sched_ctx_id, int *
 
 	int masters[nworkers];
 	int w;
-	struct _starpu_worker *worker = NULL;
 	for(w = 0; w < nworkers; w++)
 	{
 		int workerid = workerids[w];

src/datawizard/interfaces/void_interface.c

@@ -140,7 +140,7 @@ static int dummy_copy(void *src_interface STARPU_ATTRIBUTE_UNUSED,
 	return 0;
 }
 
-static ssize_t describe(void *data_interface, char *buf, size_t size)
+static ssize_t describe(void *data_interface STARPU_ATTRIBUTE_UNUSED, char *buf, size_t size)
 {
 	return snprintf(buf, size, "0");
 }

src/datawizard/memory_nodes.c

@@ -132,6 +132,7 @@ void _starpu_memory_node_get_name(unsigned node, char *name, int size)
 		prefix = "SCC_shared";
 		break;
 	case STARPU_UNUSED:
+		prefix = "unknown";
 		STARPU_ASSERT(0);
 	}
 	snprintf(name, size, "%s %u\n", prefix, descr.devid[node]);

src/drivers/cuda/driver_cuda.c

@@ -494,8 +494,7 @@ int _starpu_cuda_driver_init(struct _starpu_worker_set *worker_set)
 
 	for (i = 0; i < worker_set->nworkers; i++)
 	{
-		struct _starpu_worker *worker = &worker_set->workers[i];
-		_STARPU_TRACE_WORKER_INIT_END(worker->workerid);
+		_STARPU_TRACE_WORKER_INIT_END(worker_set->workers[i].workerid);
 	}
 
 	/* tell the main thread that this one is ready */

src/drivers/driver_common/driver_common.c

@@ -33,7 +33,7 @@
 #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, struct timespec *codelet_start, int rank, int profiling)
+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)
 {
 	struct starpu_task *task = j->task;
 	struct starpu_codelet *cl = task->cl;

src/drivers/opencl/driver_opencl.c

@@ -568,7 +568,6 @@ static void _starpu_opencl_stop_job(struct _starpu_job *j, struct _starpu_worker
 int _starpu_opencl_driver_init(struct _starpu_worker *args)
 {
 	int devid = args->devid;
-	int workerid = args->workerid;
 
 	_starpu_worker_start(args, _STARPU_FUT_OPENCL_KEY);
 
@@ -599,7 +598,7 @@ int _starpu_opencl_driver_init(struct _starpu_worker *args)
 
 	_STARPU_DEBUG("OpenCL (%s) dev id %d thread is ready to run on CPU %d !\n", devname, devid, args->bindid);
 
-	_STARPU_TRACE_WORKER_INIT_END(workerid);
+	_STARPU_TRACE_WORKER_INIT_END(args->workerid);
 
 	/* tell the main thread that this one is ready */
 	STARPU_PTHREAD_MUTEX_LOCK(&args->mutex);

src/sched_policies/locality_work_stealing_policy.c

@@ -288,15 +288,15 @@ static void lws_add_workers(unsigned sched_ctx_id, int *workerids,unsigned nwork
 		for(;;)
 		{
 			neighbour = (struct starpu_tree*)it.value;
-			int workerids[STARPU_NMAXWORKERS];
-			int nworkers = _starpu_worker_get_workerids(neighbour->id, workerids);
+			int neigh_workerids[STARPU_NMAXWORKERS];
+			int neigh_nworkers = _starpu_worker_get_workerids(neighbour->id, neigh_workerids);
 			int w;
-			for(w = 0; w < nworkers; w++)
+			for(w = 0; w < neigh_nworkers; w++)
 			{
-				if(!it.visited[workerids[w]] && workers->present[workerids[w]])
+				if(!it.visited[neigh_workerids[w]] && workers->present[neigh_workerids[w]])
 				{
-					ws->proxlist[workerid][cnt++] = workerids[w];
-					it.visited[workerids[w]] = 1;
+					ws->proxlist[workerid][cnt++] = neigh_workerids[w];
+					it.visited[neigh_workerids[w]] = 1;
 				}
 			}
 			if(!workers->has_next(workers, &it))

src/sched_policies/work_stealing_policy.c

@@ -1,6 +1,6 @@
 /* StarPU --- Runtime system for heterogeneous multicore architectures.
  *
- * Copyright (C) 2010-2013  Université de Bordeaux 1
+ * Copyright (C) 2010-2014  Université de Bordeaux 1
  * Copyright (C) 2010, 2011, 2012, 2013  Centre National de la Recherche Scientifique
  * Copyright (C) 2011, 2012  INRIA
  *
@@ -258,9 +258,7 @@ static inline unsigned select_worker(unsigned sched_ctx_id)
 }
 
 
-#ifdef STARPU_DEVEL
-#warning TODO rewrite ... this will not scale at all now
-#endif
+/* Note: this is not scalable work stealing,  use lws instead */
 static struct starpu_task *ws_pop_task(unsigned sched_ctx_id)
 {
 	struct _starpu_work_stealing_data *ws = (struct _starpu_work_stealing_data*)starpu_sched_ctx_get_policy_data(sched_ctx_id);

src/worker_collection/worker_list.c

@@ -1,6 +1,6 @@
 /* StarPU --- Runtime system for heterogeneous multicore architectures.
  *
- * Copyright (C) 2013  Université de Bordeaux 1
+ * Copyright (C) 2013-2014  Université de Bordeaux 1
  * Copyright (C) 2012-2013  Centre National de la Recherche Scientifique
  * Copyright (C) 2011-2013  INRIA
  *
@@ -173,7 +173,7 @@ static int list_remove(struct starpu_worker_collection *workers, int worker)
 static void _init_workers(int *workerids)
 {
 	unsigned i;
-	int nworkers = starpu_worker_get_count();
+	unsigned nworkers = starpu_worker_get_count();
 	for(i = 0; i < nworkers; i++)
 		workerids[i] = -1;
 	return;