perf stat: Fix event grouping on forked task
Namhyung Kim [Fri, 16 Mar 2012 08:42:19 +0000 (17:42 +0900)]
When event group is enabled for forked task (i.e. no target task was
specified) all events were disabled and marked ->enable_on_exec.
However they are not counted at all since only group leader will be
enabled on exec actually. So the result looked like below:

 $ ./perf stat --group -- sleep 1

 Performance counter stats for 'sleep 1':

          0.554926 task-clock                #    0.001 CPUs utilized
     <not counted> context-switches
     <not counted> CPU-migrations
     <not counted> page-faults
     <not counted> cycles
   <not supported> stalled-cycles-frontend
   <not supported> stalled-cycles-backend
     <not counted> instructions
     <not counted> branches
     <not counted> branch-misses

       1.001228093 seconds time elapsed

Fix it by disabling group leader only.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1331887340-32448-1-git-send-email-namhyung.kim@lge.com
Signed-off-by: Namhyung Kim <namhyung.kim@lge.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools/perf/builtin-stat.c

index ea40e4e..c941bb6 100644 (file)
@@ -296,7 +296,7 @@ static int create_perf_stat_counter(struct perf_evsel *evsel,
        if (system_wide)
                return perf_evsel__open_per_cpu(evsel, evsel_list->cpus,
                                                group, group_fd);
-       if (!target_pid && !target_tid) {
+       if (!target_pid && !target_tid && (!group || evsel == first)) {
                attr->disabled = 1;
                attr->enable_on_exec = 1;
        }