Thanks Benny for all this information. Well at least now I have something absolutely optimized :) (but as I already found something really similar on the wiki almost all flags was already well set). But unfortunately doesn't solve completely the issue - even if CPU usage is now really reduced (about 20% even when it is going wrong and with iLBC). There is maybe something to see with floating point build (thanks to Samuel for the idea, I'll check that) or with my audio driver or the android streams. But still a little bit amazing : when it works even if I stop/start audio stream or hold/reinvite it still works then. And the same consistancy can be observed when it doesn't work (choppy sound). I'll do more profiling on my current work and keep you informed if something new found and that has something to do with pjsip. Thanks again ! R?gis P.S. : any news about my feature request on penh (voice enhancer) settings within psjua or/and config site? I still think that it could be really interesting 2010/8/18 Benny Prijono <bennylp at teluu.com> > 2010/8/18 R?gis Montoya <r3gis.3r at gmail.com>: > > What is strange is that linphone that has also an android version and use > > the same ilbc implementation and it works fine : > > The begin is choppy but after 3 seconds things become as fine as it is > using > > pjsip when it works properly. > > > > Is that possible that when things are going wrong at the beggining (when > CPU > > is the more used) it goes in a state where pjsip try to correct things > which > > consume more cpu and so it goes worse. > > The only thing I can imagine is the log, which the jitter buffer could > write too aggresively on certain circumstances, which may take out > precious processing power since writing to file is a blocking > operation. > > Other than that, have a look at config_site_sample.h for suitable > settings for mobile devices, and also > https://trac.pjsip.org/repos/wiki/FAQ#cpu on optimizing PJSIP in > general. > > > Probably, when things goes wrong, linphone just ignore this (and doesn't > try > > to restablish things) and as then cpu is less used everything is better > > after 2 or 3 seconds of choppy sound. > > > > I repeat but it is really strange that with pjsip sometimes it works > > perfectly (and with a better quality than linphone). > > Is there a pjsua method I can test that (while in call) reset the rtp > stream > > reader/writer? (I tried to stop / restart the stream but doesn't seems to > > affect network data processing, just the audio one) > > > > Doing re-INVITE or UPDATE should restart the stream. > > -benny > > _______________________________________________ > Visit our blog: http://blog.pjsip.org > > pjsip mailing list > pjsip at lists.pjsip.org > http://lists.pjsip.org/mailman/listinfo/pjsip_lists.pjsip.org > -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.pjsip.org/pipermail/pjsip_lists.pjsip.org/attachments/20100820/beb16498/attachment.html>
