On Apr 01, 2013, at 16:02, Nicolas George wrote: > Le duodi 12 germinal, an CCXXI, "René J.V. Bertin" a écrit : >> I may be mistaken, but I think that the higher your timer precision, the >> lower the jitter such 'force-majeur' events produce (or, the jitter is >> proportional to your timer's precision). Right? > > I believe you are wrong: you mix timer precision and time slice. > > On a NOHZ Linux x86_64 kernel and an idle system, the precision of nanosleep > seems to be 100 µs, but if your system is busy, the process will be > preempted for a full time slice, and that is much larger, above 500 µs. > > Regards,
In any case we seem to agree that it all depends on how the timer is implemented, and of course that goes both ways. Your nanosleep function that has a precision 100000x worse than what one might expect seems to be a good case in point. (What's a NOHZ kernel?) _______________________________________________ Libav-user mailing list [email protected] http://ffmpeg.org/mailman/listinfo/libav-user
