Hi, I find this problem completely reproducible by running the stress program (from the stress package).
Do this: (1) Open a terminal, run mplayer SOMEaudioFILE.ogg Audio output is typically okay at this point (at least for me). (2) Open another terminal, run stress -c [noOFtotalCPUcores] In my case, that's 2. When you do this, the stuttering begins (and is very, very bad). Not only that, all process (apart from stress) perform very poorly. It seems like the scheduler is (almost completely) ignoring all other processes in favour of the one that is already using 100% CPU. It seems, though, that if the number of stress-spawned processes is not the same as the number of cores, e.g. 1 or 3, then the problem is not nearly as bad, even though the processes are stressing the CPU. For example, when I run stress -c 3, CPU usage on both cores is in total 100%, yet the sound output is much better and other processes are much more responsive, however a stutter can be heard now and then. When you crank the number of stress processes to something ridiculous, say 30, sound output is even better! I might have heard a single stutter during this test, I'm not even sure (while 2 cores are running at 100% total each, I think this is pretty good). UI responsiveness is somewhat worse, though. Maybe this is what is causing the stutter under normal use: Some app spikes in CPU usage for a short period, causing the audio process(es) to be stalled, thus the stuttering. The behaviour affects all process, but audio being what it is, that is where it's readily obvious that something is wrong. I probably should say I was under the impression that under general use the stuttering was a lot worse when I first encountered it (when I first tried 2.6.26) than it is now, so something has improved. If you do any of the above in 2.6.25 there is no problem whatsoever. No stuttering. Even the UI is much more responsive, for any number of processes, than in 2.6.26. I'd go so far as to state that 2.6.25 is more responsive under load than 2.6.26, which sounds a lot like a regression to me. Info: 00:06.1 Audio device: nVidia Corporation MCP55 High Definition Audio (rev a2) Sep 27 09:33:53 phi kernel: [ 0.000000] Linux version 2.6.26-1-amd64 (Debian 2.6.26-5) ([EMAIL PROTECTED]) (gcc version 4.1.3 20080623 (prerelease) (Debian 4.1.2-23 )) #1 SMP Wed Sep 10 15:31:12 UTC 2008 as compared to: Sep 27 10:42:02 phi kernel: [ 0.000000] Linux version 2.6.25-2-amd64 (Debian 2.6.25-7) ([EMAIL PROTECTED]) (gcc version 4.1.3 20080623 (prerelease) (Debian 4.1.2-23) ) #1 SMP Mon Jul 14 11:05:23 UTC 2008 Peace, Brendon
signature.asc
Description: This is a digitally signed message part.
