linux-perf-users.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* duplicate cache misses counter labels with perf-stat
@ 2012-09-05 15:21 Joel Uckelman
  2012-09-05 17:51 ` [PATCH] perf tools: Fix cache event name generation Jiri Olsa
  0 siblings, 1 reply; 2+ messages in thread
From: Joel Uckelman @ 2012-09-05 15:21 UTC (permalink / raw)
  To: linux-perf-users

When I do this with perf 3.5.2,

  perf stat -x, -e
L1-dcache-loads,L1-dcache-load-misses,L1-dcache-stores,L1-dcache-store-misses
ls

I'm seeing this output from perf:

  697171,L1-dcache-loads

  28012,L1-dcache-misses

  455691,L1-dcache-stores

  11791,L1-dcache-misses

So, I'm getting two results labeled "L1-dcache-misses" (with different
values!) and none labeled "L1-dcache-store-misses". It looks to me
like one of the "L1-dcache-misses" is mislabeled.

There appears to be a general problem with all of the *-misses
counters, as a similar thing happens with, e.g., LLC-load-misses and
LLC-store-missses, and all of the others I've checked.

Is this a bug, or am I misunderstanding how perf-stat is supposed to work?

^ permalink raw reply	[flat|nested] 2+ messages in thread

* [PATCH] perf tools: Fix cache event name generation
  2012-09-05 15:21 duplicate cache misses counter labels with perf-stat Joel Uckelman
@ 2012-09-05 17:51 ` Jiri Olsa
  0 siblings, 0 replies; 2+ messages in thread
From: Jiri Olsa @ 2012-09-05 17:51 UTC (permalink / raw)
  To: Joel Uckelman
  Cc: linux-perf-users, Arnaldo Carvalho de Melo, Peter Zijlstra,
	Ingo Molnar, Paul Mackerras, Corey Ashford, Frederic Weisbecker

On Wed, Sep 05, 2012 at 05:21:10PM +0200, Joel Uckelman wrote:
> When I do this with perf 3.5.2,
> 
>   perf stat -x, -e
> L1-dcache-loads,L1-dcache-load-misses,L1-dcache-stores,L1-dcache-store-misses
> ls
> 
> I'm seeing this output from perf:
> 
>   697171,L1-dcache-loads
> 
>   28012,L1-dcache-misses
> 
>   455691,L1-dcache-stores
> 
>   11791,L1-dcache-misses
> 
> So, I'm getting two results labeled "L1-dcache-misses" (with different
> values!) and none labeled "L1-dcache-store-misses". It looks to me
> like one of the "L1-dcache-misses" is mislabeled.
> 
> There appears to be a general problem with all of the *-misses
> counters, as a similar thing happens with, e.g., LLC-load-misses and
> LLC-store-missses, and all of the others I've checked.
> 
> Is this a bug, or am I misunderstanding how perf-stat is supposed to work?

bug ;) please try attached patch

thanks,
jirka


---
If the event name is specified with all 3 components, the last one
overwrites the previous one during the name composing within the
parse_events_add_cache function.

Fixing this by properly adjusting the string index.

Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
---
 tools/perf/util/parse-events.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index b246303..a031ee1 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -308,7 +308,7 @@ int parse_events_add_cache(struct list_head **list, int *idx,
 	for (i = 0; (i < 2) && (op_result[i]); i++) {
 		char *str = op_result[i];
 
-		snprintf(name + n, MAX_NAME_LEN - n, "-%s\n", str);
+		n += snprintf(name + n, MAX_NAME_LEN - n, "-%s", str);
 
 		if (cache_op == -1) {
 			cache_op = parse_aliases(str, perf_evsel__hw_cache_op,
-- 
1.7.11.4

^ permalink raw reply related	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2012-09-05 17:52 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2012-09-05 15:21 duplicate cache misses counter labels with perf-stat Joel Uckelman
2012-09-05 17:51 ` [PATCH] perf tools: Fix cache event name generation Jiri Olsa

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).