Subject: Re: [linux-audio-dev] more re csound -L
From: est_AT_hyperreal.org
Date: ti syys 07 1999 - 09:49:43 EDT
Paul Winkler discourseth:
> est_AT_hyperreal.org wrote:
> >
> > I'm getting good results. Interested parties should see
> > www.hyperreal.org/~est/csound/csoundL.tar.gz.
>
> Thanks a lot! This definitely answers some of my questions.
>
> I tried changing that last loop to run nothing but lots() 4 times, then
> run that loop 16 times. On my system, having 30 events in lots() doesn't
> work so well... it starts out sounding good, but within 5 seconds timing
> has gone haywire and notes are breaking down into a blur of successive
> notes...
That's with the single call to os.write() per call to lots? Hmm, I
don't have that problem at all. Maybe you should try it with
unofficial-csound-3.57.0.1d-linux.
> [ with another method] I can only get about n = 12 before timing starts to go
> bad. Though it doesn't sound like individual "notes" (calls to lots())
> are blurred; rather, they are misplaced in time. (Blurring happens
> quickly with much higher values of n.)
[also]
> You would think that for a given n, lotsn() would fare better than
> lotsmore() since it only ever calls os.write() once.
> But I'm not finding much difference...
It's partly a matter of reliability. Someone else might run *between*
your calls to write.
Anyhow, to really isolate these problems, you'd need to run both
processes as real-time processes, apply the real-time patches, and
upgrade your HZ (..and possibly also change the SCHED_RR
quantum..yikes! i haven't looked into that question yet.). I hadn't
done *any* of that and got good (though no doubt unreliable) results,
but lacking any of those can easily cause some of the problems you
described.
Eric
This archive was generated by hypermail 2b28 : pe maalis 10 2000 - 07:27:11 EST