Small improvements for Papi counters

* accept comma as separator of events to monitor
* warn is event registration failed because of unavailable component
* doc: explain how to get available counters (the given link isn't
  online anymore)
* doc: tell about the perf_event_paranoid kernel parameter

doc/doxygen/chapters/380_offline_performance_tools.doxy

@@ -586,19 +586,31 @@ $ starpu_paje_sort paje.trace
 \section PapiCounters PAPI counters
 
 Performance counter values could be obtained from the PAPI framework if
-<c>./configure</c> detected the libpapi. One has to set the \ref STARPU_PROFILING
-environment variable to 1 and then specify which events to record with the
-\ref STARPU_PROF_PAPI_EVENTS environment variable. For instance:
+<c>./configure</c> detected the libpapi.
+
+In Debian, packages <c>libpapi-dev</c> and <c>libpapi5.7</c> provide required
+files.  Package <c>papi-tools</c> contains a set of useful tools, for example
+<c>papi_avail</c> to see which counters are available.
+
+To be able to use Papi counters, one may need to reduce the level of the kernel
+parameter <c>kernel.perf_event_paranoid</c> to at least 2. See
+https://www.kernel.org/doc/html/latest/admin-guide/perf-security.html for the
+security impact of this parameter.
+
+Then one has to set the \ref STARPU_PROFILING environment variable to 1 and
+specify which events to record with the \ref STARPU_PROF_PAPI_EVENTS
+environment variable. For instance:
 
 \verbatim
 export STARPU_PROFILING=1 STARPU_PROF_PAPI_EVENTS="PAPI_TOT_INS PAPI_TOT_CYC"
 \endverbatim
 
+The comma can also be used to separate events to monitor.
+
 In the current simple implementation, only CPU tasks have their events measured
-and require CPUs that support the PAPI events. All events that PAPI support are
-available from their documentation (https://icl.cs.utk.edu/projects/papi/wiki/PAPIC:Preset_Event_Definitions).
-It is important to note that not all events are available on all systems, and
-general PAPI recommendations should be followed.
+and require CPUs that support the PAPI events. It is important to note that not
+all events are available on all systems, and general PAPI recommendations
+should be followed.
 
 The counter values can be accessed using the profiling interface:
 \code{.c}

src/profiling/profiling.c

@@ -46,6 +46,7 @@ static struct timespec executing_start_date[STARPU_NMAXWORKERS];
 #ifdef STARPU_PAPI
 static int papi_events[PAPI_MAX_HWCTRS];
 static int papi_nevents = 0;
+static int warned_component_unavailable = 0;
 #endif
 
 /* Store the busid of the different (src, dst) pairs. busid_matrix[src][dst]
@@ -160,7 +161,7 @@ void _starpu_profiling_init(void)
 		conf_papi_events = starpu_getenv("STARPU_PROF_PAPI_EVENTS");
 		if (conf_papi_events != NULL)
 		{
-			while ((papi_event_name = strtok_r(conf_papi_events, " ", &conf_papi_events)))
+			while ((papi_event_name = strtok_r(conf_papi_events, " ,", &conf_papi_events)))
 			{
 				_STARPU_DEBUG("Loading PAPI Event:%s\n", papi_event_name);
 				retval = PAPI_event_name_to_code ((char*)papi_event_name, &papi_events[papi_nevents]);
@@ -188,7 +189,12 @@ void _starpu_profiling_papi_task_start_counters(struct starpu_task *task)
 		PAPI_create_eventset(&profiling_info->papi_event_set);
 		for(int i=0; i<papi_nevents; i++)
 		{
-			PAPI_add_event(profiling_info->papi_event_set, papi_events[i]);
+			int ret = PAPI_add_event(profiling_info->papi_event_set, papi_events[i]);
+			if (ret == PAPI_ECMP_DISABLED && !warned_component_unavailable)
+			{
+				_STARPU_MSG("Error while registering Papi event: Component containing event is disabled. Try running `papi_component_avail` to get more information.\n");
+				warned_component_unavailable = 1;
+			}
 			profiling_info->papi_values[i]=0;
 		}
 		PAPI_reset(profiling_info->papi_event_set);