Comment 7 for bug 579117

Now that I've watched the messages file enough, I see there's a pattern;
I get several successive underruns fairly quickly with the "wakeup
watermark" increased from 30ms through 60ms, then things level off for
awhile.

Anyway, after the first underrun, here's what I see in pulseaudio:

scheduler: waiting for cpu 19.8ms 2.5%
userspace lock contention 4.0ms 0.1%
waiting for event (poll) 1.5ms 0.0%

ok I've unfrozen latencytop and now I'm waiting for another underrun...

ok there's one, now we have given up on increasing wakeup watermark and
we are increasing minimal latency to 1.00 ms

here's what I see now in pulseaudio:

scheduler: waiting for cpu 2.6ms 7.9%
waiting for event (poll) 2.1ms 0.1%

another unfreeze, another underrun where minimal latency is increased to
2.00ms

in pulseaudio, now we only have

scheduler: waiting for cpu 2.3ms 11.2%

ok no new underruns, but the last automatic refresh in latencytop showed
a big value - 22.5 ms - in Scheduler: waiting for cpu... hmmmm

and then there was an underrun

now as I watch, I see typically 3.0ms...

another underrun, the max is showing at 4.6ms... now it's 1.9ms

I've had 10 underruns since starting rhythmbox up 20 minutes ago.

Does any of this help?

On Wed, 2010-05-19 at 00:52 +0000, David Henningsson wrote:

> Oh, Latencytop has got a new UI for Lucid, nice :-) Although it is still
> missing decent documentation :-(
>
> Anyway, I guess the most important thing would be to look at the
> "pulseaudio" process after an underrun and see if anything seems strange
> there, or if you get anything above 20 ms.
>

--
Regards,

Chris Hermansen · mailto:<email address hidden>