perf: Use generic sample reordering in perf sched
Frederic Weisbecker [Fri, 23 Apr 2010 22:29:40 +0000 (00:29 +0200)]
Use the new generic sample events reordering from perf sched,
this drops the need of multiplexing the buffers on record time,
improving the scalability of perf sched.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Tom Zanussi <tzanussi@gmail.com>

tools/perf/builtin-sched.c

index 09ddc8e..94453f1 100644 (file)
@@ -1651,9 +1651,10 @@ static int process_lost_event(event_t *event __used,
 }
 
 static struct perf_event_ops event_ops = {
-       .sample = process_sample_event,
-       .comm   = event__process_comm,
-       .lost   = process_lost_event,
+       .sample                 = process_sample_event,
+       .comm                   = event__process_comm,
+       .lost                   = process_lost_event,
+       .ordered_samples        = true,
 };
 
 static int read_events(void)
@@ -1850,7 +1851,6 @@ static const char *record_args[] = {
        "record",
        "-a",
        "-R",
-       "-M",
        "-f",
        "-m", "1024",
        "-c", "1",