On Tuesday 18 April 2006 02:54, Loki Davison was like:
> One note: Based on a comment Lee Revel made a
few days ago, I think
> on this list, about the standard kernel giving good realtime results I
> built 2.6.16-gentoo-r2 on Saturday. It's now up two days with no
> problems. I'm running at 128/2, and even 64/2 as shown below) in Jack
> with no xruns so far although I've got a cold and am using it very
> lightly right now. Anyway, I agree with Lee (so far!) that even the
> current 2.6.16 series is giving much better performance than older
> kernels even without Ingo's patches. IF it matters I use the
> realtime-lsm module and not the othe PAM based stuff.
I just bit the bullet and decided to have a go at rolling my own kernel for
the first time, following the DeMuDi HOWTO
http://demudi.agnula.org/wiki/Low-latencyKernelBuildingHowto and a certain
amount of advice from Antonio and well, 5 builds later I seem to have a
working kernel! I´m using the patched libpam modules from Ubuntu (heh ;) with
2.6.16.5 so far so good. Ardour-0.99 seems happy, with a really low dsp.
Certainly as good as I´ve had with earlier -rt patched kernel packages from
DeMuDi.
I´m wondering whether setting negative nice values
in /etc/security/limits.conf is necessary or even a good idea. I shall have
to play around and fine-tune things a bit yet.
Now I have started I may as well build an -rt patched version to test the
difference. Does it conflict with PAM rtlimits?
--
cheers,
tim hall
http://glastonburymusic.org.uk/tim