On Tue, Jan 29, 2013 at 12:25:31PM +0100, Andreas Hollmann wrote:
> Hi,
>
> I've noticed an odd behavior of the perf event subsystem.
>
> If the flag "disabled" in struct perf_event_attr is set to 0, the
> counter should be automatically started after the perf_event_open
> system call (perf/design.txt).
>
> This is only working if the first perf counter is of type hw or all
> counters in the group are sw counters. If a sw counter comes first
> and the group also contains hw counters, the read call shows that the
> first counter is started, but not counting. All other counters are neither
> started nor counting.
>
> Value: 0 Enabled: 7599 Running: 6706
> Value: 0 Enabled: 0 Running: 0
> Value: 0 Enabled: 0 Running: 0
hm, I'm getting following stats when running perf stat in similar scenario:
[jolsa@dhcp-26-214 perf]$ ./perf stat -e '{task-clock,cpu-clock,instructions}'
kill
usage: kill [ -s signal | -p ] [ -a ] pid ...
kill -l [ signal ]
Performance counter stats for 'kill':
1.039561 task-clock # 0.479 CPUs utilized
<not counted> cpu-clock
2,281,726 instructions # 0.00 insns per cycle
0.002168502 seconds time elapsed
...
seems wrong ;-) will check
thanks,
jirka
--
To unsubscribe from this list: send the line "unsubscribe linux-perf-users" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
