[PD] cpu load question

Frank Barknecht fbar at footils.org
Sat Jan 11 00:44:37 CET 2003

Pall Thayer hat gesagt: // Pall Thayer wrote:

> I just can't figure this one out. I've got a linux machine that's
> been running two seperate pd patches for the past 3 or 4 months with
> no problems. When I look at 'top' I see that each of them (they're
> running as seperate instances of pd with no gui, somewhere in the
> background) are using from 2 to 8 percent of the cpu. Now I'm
> working on adding a third one, it's actually not much different than
> the first one but when I run it, it uses 50 to 70 percent of the
> cpu! Both of these patches, the 1st and 3rd, are running two
> seperate sequencers, a mixture of some other minor stuff and sending
> the whole shebang to an icecast Mp3 server. But the two sequencers
> are created quite differently. In the new one (the one that eats up
> the cpu) I step through arrays for the sequences. On the first one,
> I didn't know how to use arrays and managed to create a very
> complicated system that looks like it MUST do more than play an
> eight step sequence (but it doesn't). Anyway, when I figured out the
> arrays I thought, "Hey, that should make it more cpu friendly. But
> of course, that might not necessarily be the problem. In the new
> one, I'm also running two envelope followers and the flashserver
> object. I've gone through and tried removing bits and pieces to see
> if anything will lighten the load but nothing appears to make any
> significant difference, not even when I remove the env~ objects. I'm
> frustrated.

Well, I suspect, this is difficult to solve without seeing the
patches. I don't know, how you did it, but I can't imagine, that the
array-seqeuencer is faulty here. I use lots of [GUI heavy] sequencers,
that store steps in arrays, and they don't have much cpu load (you
might look at them at footils.org, the family is called "sseq".) Also
env~ isn't that CPU hungry, normally. I don't know flashserver,

 Frank Barknecht                               _ ______footils.org__

More information about the Pd-list mailing list