[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20241218084656.609764-1-yeoreum.yun@arm.com>
Date: Wed, 18 Dec 2024 08:46:56 +0000
From: Yeoreum Yun <yeoreum.yun@....com>
To: peterz@...radead.org,
mingo@...hat.com,
acme@...nel.org,
amhyung@...nel.org,
mark.rutland@....com,
alexander.shishkin@...ux.intel.com,
jolsa@...nel.org,
irogers@...gle.com,
adrian.hunter@...el.com,
kan.liang@...ux.intel.com,
james.clark@...aro.org
Cc: linux-perf-users@...r.kernel.org,
linux-kernel@...r.kernel.org,
nd@....com,
Yeoreum Yun <yeoreum.yun@....com>
Subject: [PATCH 1/1] events/core: fix error run/enable ratio display
when runs below command (with core pmu) sometime perf prints error
ratio for each perf.
sudo ./perf stat -vvv -e armv8_pmuv3_0/event=0x08/ -e armv8_pmuv3_1/event=0x08/ -- stress-ng --pthread=2 -t 2s
...
armv8_pmuv3_0/event=0x08/: -1: 1081702126 2289429840 2174835740
armv8_pmuv3_1/event=0x08/: -1: 794080238 1950025700 847648440
armv8_pmuv3_0/event=0x08/: 1138698008 2289429840 2174835740
armv8_pmuv3_1/event=0x08/: 1826791390 1950025700 847648440
Performance counter stats for 'stress-ng --pthread=2 -t 2s':
1,138,698,008 armv8_pmuv3_0/event=0x08/ (94.99%)
1,826,791,390 armv8_pmuv3_1/event=0x08/ (43.47%)
This happens because of missing total_enable_time of inactivate event.
it's enabled time doesn't update at the event_sched_out() because state
isn't PERF_EVENT_STATE_ACTIVE but PERF_EVENT_STATE_INACTIVE.
For example, Suppose there're two events:
e0: with pmu0 which supports only cpu0
e1: with pmu1 which supports only cpu1.
then
CPU0 CPU1 CPU0
| in | out | | in | out | | in | out (exit) |
------------------------------------------------------------------------------
ctx->time | 0 | t1 | | t1 | t1 + t2 | | t1 + t2 | t1 + t2 + t3 |
------------------------------------------------------------------------------
e0->ena | 0 | t1 | | t1 | t1 | | t1 + t2 | t1 + t2 + t3 |
------------------------------------------------------------------------------
e0->run | 0 | t1 | | t1 | t1 + t2 | | t1 + t2 | t1 + t3 |
------------------------------------------------------------------------------
e1->ena | 0 | 0 | | t1 | t1 + t2 | | t1 + t2 | t1 + t2 |
------------------------------------------------------------------------------
e1->run | 0 | 0 | | 0 | t2 | | t1 + t2 | t2 |
------------------------------------------------------------------------------
at the CPU0, state of e1 is PERF_EVENT_STATE_INACTIVE since it doesn't
support on CPU0. So when event_sched_out() it failed to update its
total_enable_time and it makes error print of run/ena ratio.
To address this, update total_enable_time in event_sched_out() when event state
is PERF_EVENT_STATE_INACTIVE.
After this patch, this problem is gone like:
sudo ./perf stat -vvv -e armv8_pmuv3_0/event=0x08/ -e armv8_pmuv3_1/event=0x08/ -- stress-ng --pthread=2 -t 2s
...
armv8_pmuv3_0/event=0x08/: -1: 10484528214 32157963940 21898169000
armv8_pmuv3_1/event=0x08/: -1: 7155819373 32157963940 10259794940
armv8_pmuv3_0/event=0x08/: 15396770398 32157963940 21898169000
armv8_pmuv3_1/event=0x08/: 22428964974 32157963940 10259794940
Performance counter stats for 'stress-ng --pthread=2 -t 10s':
15,396,770,398 armv8_pmuv3_0/event=0x08/ (68.10%)
22,428,964,974 armv8_pmuv3_1/event=0x08/ (31.90%)
Signed-off-by: Yeoreum Yun <yeoreum.yun@....com>
---
kernel/events/core.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 065f9188b44a..f40cca02636d 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2311,6 +2311,11 @@ event_sched_out(struct perf_event *event, struct perf_event_context *ctx)
WARN_ON_ONCE(event->ctx != ctx);
lockdep_assert_held(&ctx->lock);
+ if (event->state == PERF_EVENT_STATE_INACTIVE) {
+ perf_event_update_time(event);
+ return ;
+ }
+
if (event->state != PERF_EVENT_STATE_ACTIVE)
return;
--
LEVI:{C3F47F37-75D8-414A-A8BA-3980EC8A46D7}
Powered by blists - more mailing lists