On Wed, 2007-09-19 at 21:48 -0400, lanas wrote: > Le Dimanche, 16 Septembre 2007 14:54:52 +0100, > David Haggett <david@xxxxxxxxxxxxxxxxxxx> a écrit : > > > Until recently I have been running SUSE 9.3 x86_64 with a > > self-compiled realtime kernel (using the realtime-lsm). I have now > > upgraded to openSUSE 10.2 because I was having difficulties > > installing the latest versions of certain applications. > > > > I'm very happy with the distribution generally, but I'd now like to > > get my audio performance back to what it was. Via the installation > > and set-up forum at jacklab, I have been directed to realtime kernel, > > kernel-source and PAM packages for x86_64 at: > > This raises an interesting question. How come I can use the plain > Fedora Core 6 kernel (and not the real-time CCRMA kernel) to do audio > works on a x86_64 dual core with 4 GB RAM without any problems ? Is it > because I'm not doing much ? I mean, I do not record a chamber > orchestra in real time while jiggling the screen with beryl, > compiling OpenOffice and letting goddamned beagle run loose all at the > same time ? And leaving Konqueror open with one of those pages that > mysteriously drains CPU cycles to make good measure. > > Or is it the dual core ? Or the RAM ? Hard to say exactly why. The realtime preemption patches help when you are doing audio at low latencies. What is low? Hmmm, say 5 msec or below (2x128, for example, or even 2x64), but that's not a hard number. Do you need such low latencies? It depends on what kind of work you do. Latencies depend also on the hardware you have, how it is configured, etc, etc. Also on how critical you are of the performance you get from the system. Do you run Jack? Does it ever signal that there was an xrun with the stock Fedora kernel? It could be that you are getting some xruns and you have not noticed yet. And the problem, if you think about it, is that you don't know what could cause one - you can always try to be on the safe side with what you do but what if one tiny xrun happens in the middle of some critical work? > Or is it because the Fedora team already incorporates real-time > mechanism into their kernels ? AFAIK it is compiled with PREEMPT_VOLUNTARY which is better than no preemption at all. And if it is a very recent kernel it will have many things in it that will make it better latency-wise. -- Fernando _______________________________________________ Linux-audio-user mailing list Linux-audio-user@xxxxxxxxxxxxxxxxxxxx http://lists.linuxaudio.org/mailman/listinfo.cgi/linux-audio-user
