Re: [PATCH 12/15] perf tools: allow non-matching sample types

[David Ahern]

On 6/25/13 5:04 PM, David Ahern wrote:
> On 6/25/13 10:03 AM, Stephane Eranian wrote:
> > > Stephane: are you looking at allowing sample_types per event?
> > > >
> > Yes, this is what I need. I have a kernel patch to do this. I don't
> > know how to update perf to handle it correctly. So maybe you can
> > help. My patch is useful to drastically reduce the size of the perf.data
> > file in case we use the branch-stack with lots of events which is
> > what our Gooda tool would like to do.
>
> Refreshing my memory on the root problem here. It's a chicken-and-egg
> problem: we need the id in the sample to find the event (evsel) that
> generated it (perf_evlist__id2evsel). To get the id we need the
> sample_type to parse it and we want the sample_type to be per event.
>
> As I recall this is where the conversation turns to per-event data files...

That said, this hack works for me with the combined S/W-tracepoint data 
file:

$ perf record -e cs -c1 -e sched:sched_switch -a -- sleep 1
$ perf script

It basically guesses which entry in the array has the id. (Ignore the 
whitespace ugliness - dev box expands tabs).

diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 89aea20..a6824b4 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -849,7 +849,20 @@ int perf_evlist__start_workload(struct perf_evlist 
*evlist)
 int perf_evlist__parse_sample(struct perf_evlist *evlist, union 
perf_event *event,
 			      struct perf_sample *sample)
 {
-	struct perf_evsel *evsel = perf_evlist__first(evlist);
+	struct perf_evsel *evsel;
+    const u64 *array = event->sample.array;
+    u64 id;
+    int n;
+
+    for (n = 0; n < 4; ++n) {
+        id = array[n];
+        evsel = perf_evlist__id2evsel(evlist, id);
+        if (evsel)
+            break;
+    }
+    if (evsel == NULL)
+        evsel = perf_evlist__first(evlist);
+
 	return perf_evsel__parse_sample(evsel, event, sample);
 }

diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index ad47fb9..dbbda09 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -61,10 +61,10 @@ static int perf_session__open(struct perf_session 
*self, bool force)
 		goto out_close;
 	}

-	if (!perf_evlist__valid_sample_type(self->evlist)) {
-		pr_err("non matching sample_type");
-		goto out_close;
-	}
+//	if (!perf_evlist__valid_sample_type(self->evlist)) {
+//		pr_err("non matching sample_type");
+//		goto out_close;
+//	}

 	if (!perf_evlist__valid_sample_id_all(self->evlist)) {
 		pr_err("non matching sample_id_all");
@@ -1295,10 +1295,15 @@ int perf_session__process_events(struct 
perf_session *self,

 bool perf_session__has_traces(struct perf_session *session, const char 
*msg)
 {
-	if (!(perf_evlist__sample_type(session->evlist) & PERF_SAMPLE_RAW)) {
-		pr_err("No trace sample to read. Did you call 'perf %s'?\n", msg);
-		return false;
-	}
+    struct perf_evsel *evsel;
+
+    list_for_each_entry(evsel, &session->evlist->entries, node) {
+		if ((evsel->attr.type == PERF_TYPE_TRACEPOINT) &&
+		   !(evsel->attr.sample_type & PERF_SAMPLE_RAW)) {
+			pr_err("No trace sample to read. Did you call 'perf %s'?\n", msg);
+			return false;
+		}
+    }

 	return true;
 }

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/