Re: [RFC 3/6] perf/core: use rb-tree to sched in event groups

[David Carrillo-Cisneros]

On Thu, Jan 12, 2017 at 5:28 AM, Liang, Kan <kan.liang@xxxxxxxxx> wrote:
>
>
>> On Wed, Jan 11, 2017 at 08:31:11PM +0000, Liang, Kan wrote:
>>
>> > > Kan, in your per-cpu event list patch you mentioned that you saw a
>> > > large overhead in perf_iterate_ctx() when skipping events for other
>> CPUs.
>> > > Which callers of perf_iterate_ctx() specifically was that
>> > > problematic for? Do those callers only care about the *active* events,
>> for example?
>> >
>> > Based on my test, the large overhead was observed in perf_iterate_sb.
>> > Yes, it only cares about the *active* events.
>>
>
> Oh Sorry, my bad. My brain must not be working yesterday...
> I just re-visited the code. The *inactive* also need to be checked.
>         if (event->state < PERF_EVENT_STATE_INACTIVE)
>                 continue;
> So the perf_iterate_sb iterates through all events (both STATE_ACTIVE
> and STATE_INACTIVE).
>
> I'm not sure if it is enough to only take care of *active* events.

The patch 6/6 in this series is the one that changes perf_iterate and
does iterate over active and
inactive events, so it only skips events in OFF and ERROR state.

> Peter may comments on it.
>