[PlanetCCRMA] audio breaking and cpu load?

[Fernando Lopez-Lezcano]

On Mon, 2010-02-22 at 14:54 -0800, Oded Ben-Tal wrote:
> > What do you see if you run "top"? What is eating up cpu? (what is your
> > period size in jack?) And what happens when you start your pd patches?
> > If the load is getting close to, say, 80%, then any spikes in processing
> > will trigger xruns.
> 
> I'm starting jack with:
> /usr/bin/jackd -R -t10000 -dfirewire -r48000 -p256 -n3 -i2 -o6
> (also tried 128 & 512).
> I get this error message from jack (but only rarely)
> JackEngine::XRun: client = pure_data_0 was not run: state = 2
> JackAudioDriver::ProcessAsync Process error

That means, I think, that pd took too long to respond with samples and
missed a cycle. 

> should the delay settings in pd (under audio settings) make any difference 
> when it's working with jack or does it take all this from jack?
> 
> top shows that it's mainly jack and pd (and Xorg occasionally. I presume 
> from the gui) taking cpu time.

Most probably the firewire driver. 

> Interestingly, sometime things run smoothly - I just ran the patch 'on 
> idle' dsp on but no user activity for 3 hours with very few recorded DIO 
> errors from pd (last one 9800 seconds ago). But then again, sometimes when 
> I run the same patch it's constantly breaking.

And what would be the cpu load when you are running the patch? It is
crucial to know that (at least run "top" and report the numbers when the
patch is running). 

> I was using the cpuload gnome applet to see what's going on (didn't use 
> qjackctl to see if it makes any difference).

And what happens when you run your pd patch?

-- Fernando