[PD] curious Pd behavior

[Ivica Ico Bukvic]

Hi all,

In my recent performance, I've noticed a peculiar PD behavior. Now, before I
ask for your advice, please allow me to state that the problem has not been
as pronounced in the past as it was on this performance. Nonetheless, it was
more or less persistent. For this reason, I am asking for your thoughts on
it.

Here's the hw/sw scoop:

AMD64 3000+ laptop
RME HDSP Multiface
Linux 2.6.15 Kernel (real-time enabled, can't recall if it is preempted)
ALSA
JACK
Pd (0.38.4) -rt running MIDI/Display interface for performer/DSP
2 x JACK-RACK with 4 LADSPA plug-ins each
2 x QSynth
The entire setup consistently eats approx. 80% cpu (no spikes).

Pd also had pyext (0.1.3) and a couple of simpler externals (yes, stuff is a
bit outdated...). All audio is diffused (and partially processed) via Pd.

What happens during the runtime is two interesting issues:

1) if I start Pd by clicking on a link (no terminal) the pyext does not
output anything when DSP is on, but as soon as I turn DSP off, it spits out
backlogged data. However, when I start pd from a terminal, everything works
ok. Any ideas as to why?

2) when CPU usage climbs above 50%, Pd starts going out of sync by
initiating opening queue twice which obviously throws entire piece out of
sync. Interestingly, this only happens approx. 50% of the time (randomly).
As the CPU usage grows, it becomes worse. This is not a problem when CPU
usage is lower. Than 50% and never a problem when using built-in soundcard
(this was not a problem before, so it may very well be that it is my setup).

3) when restarting DSP, some of the routed signal going through Pd to the
output is lost at random times. This was in part a problem before as well.

So, while I am aware very much that this may be peculiar to the versions I
used, and/or possibly due to use of a custom (possibly buggy) build of a
kernel, I was wondering if anyone had an alternative explanation for this
kind of behavior.

Many thanks!

Best wishes,

Ivica Ico Bukvic, D.M.A.
Composition, Music Technology, CCTAD, and CHCI (by courtesy)
Virginia Tech
Dept. of Music - 0240
Blacksburg, VA 24061
(540) 231-1137
(540) 231-5034 (fax)
ico at vt.edu
http://www.music.vt.edu/people/faculty/bukvic/