[PlanetCCRMA] added: ALPHA version of Planet Core components for FC2

Fernando Pablo Lopez-Lezcano nando@ccrma.Stanford.EDU
Fri Jun 25 08:31:01 2004

On Fri, 2004-06-25 at 02:57, Steve Harris wrote:
> On Thu, Jun 24, 2004 at 05:12:54 -0700, Fernando Pablo Lopez-Lezcano wrote:
> > On Thu, 2004-06-24 at 12:49, Fernando Pablo Lopez-Lezcano wrote:
> > > > I played with 2.6's jackd again last night, with the LD_ASSUME_KERNEL
> > > > trick (assuming I spelt it right ;), but I still cound't get anything like
> > > > relaible realtime performance.
> > > 
> > > I think that is the correct spelling. 
> > > 
> > > I still get weird behavior and I haven't had time to do any more
> > > testing. I noticed something yesterday. I would start qjackctl, then
> > > hydrogen and a bunch of other apps (pipe hydrogen through freqtweak and
> > > jackeq). 
> > 
> > _and_, I forgot, jack-rack running gverb. 
> > And I think _that_ is the problem. 

Well, probably not _the_ problem. But at least that is what seems to be
causing the xrun explosions in my tests. Could it be that the handling
of denormals has changed in newer kernels? That is done in the kernel,
right? (I have no idea). 

> I haven't been able to get anything like that kind of stability. If I run
> AMS, then try to run jamin within a minute I will get a storm of messages,
> then jackd will deteriorate to the point where apps will no longer be able
> to connect. I had to reboot a couple of times. This is running at 2x256.

I have not been able to run jamin, the latest official release does not
even start. Jack (O'Quinn) posted a patch when I mentioned it but I did
not get a chance to try it out. 

> > Can we say "denormals"??
> > The tail end of the reverb slowly dies and eventually gets to the point
> > where it fades to almost zero. That would do it, right? And explain why
> > it happens after hydrogen is done with its run? So it is not a
> > coincidence after all...
> That does sound exactly like denormals. Are you running a P4 by any
> chance?

Yes, P4 1.7G on my laptop. 

> > Exactly what software are you using to test?
> I've only tried AMS and jamin (debugging) with the RT kernel, with the
> stock fedora one I've run a variety of stuff, it xruns occasioanlly,
> as you would expect, but no storms.

That is interesting and something I should try. I have not actually
tried to run things in the stock kernel as I would have to run them as
root, and it is not possible to build the realcap kernel module :-( How
long are the xruns you see? Should I assume that the tests in the stock
kernel were done as root with SCHED_FIFO? (jackd -R).

When you say stock kernel you mean the one originally shipped with FC2,

-- Fernando