I read:
I'm helping develop an installation using pd (linux). We are quite poor, but still manage to have the choice between using a single celeron 533 MHz processor, or a SMP system with two celerons running at 523 MHz.
I dunno how well pd behaves on a smp system but in general since you will run other processes aswell I'd go for the smp system and depending on the project lots of ram + fast disks
There are four cheap sound cards in this system to give us eight outputs.
I think these must all take their input from one pd, as we need to
this is a faq, actually discussed to death on the linux audio developer list - the bottom line is it won't work if the cards are not hardware synced (a possibility that cheap cards rarely offer) mail me privately if you want the details why.
smoothly move between them.
you could OTOH hand run 4 pd's with the 4 different cards and redundant data (tmpfs ramdisk comes to mind) and use netsend/receive to control them all from one interface
Is there a good way to take advantage of both processors in the SMP configuration (and would this provide a significant performance gain over a single processor)?
I'm not experienced enough with smp to answer that
Is Open Sound Control (OSC) the only way to do this (run a pd on each processor, with a mixing stage in one of them, and OSC providing the pipe)? Does anyone know if the overhead associated with OSC would make this a bad idea? Anything better? Any simple solutions?
apart from the ramdisk approach you could use pipe/fifo OSC netsend/receive midi loopback or a combination of the above depending again on what you try to accomplish.
regards,
x