[LAU] yoshimi creeps... 0.051

cal cal at graggrag.com
Tue Jan 19 01:00:06 EST 2010


rosea grammostola wrote:
> 
>> What other apps are you using?
> Ardour, hydrogen, non-sequencer, jack-mixer, qsynth, linuxsampler for 
> instance...
> 
> You start to think it is something specific to yoshimi/zynaddsubfx...

The way zyn works pushes the bounds of good sense in terms of actually
generating it's sound in realtime, but you gradually learn the limitations
and how to have fun within those limitations. The main difference with
yoshi is the audio/midi drivers plus some refinement (nothing radical)
to the sound generation processing cycle. So far, yoshimi does both the
drivers and the generation cycle better than original zyn, but I think
we'd all like zynaddsubfx to catch up and get it right at last. Frankly
I can't see that happening any time soon, but I am a registered cynic in
that area.

> I'll do some more testing with other apps.

And I think I can predict the results. If you're running a whole bunch of
other apps on the same system at the same time as zyn/yoshi, then the cpu
grunt available to zyn/yoshi is going to reduce accordingly, and so is the
level of sound complexity you'll be able to achieve in realtime. Currently,
yoshimi sets the priority of the midi and audio threads to 50 (or 55, can't
remember), and they're competing for cpu with everything else you've got
running SCHED_FIFO.

If you really, really, really want the sounds zyn/yoshi can produce, then
you learn to accommodate its demands and limitations in terms of resource
availability. If you can't accommodate those requirements, then you really
need to use a different synth.  

cheers.




More information about the Linux-audio-user mailing list