Re: [LAU] yoshimi falls asleep on padsynth load

From: cal <cal@email-addr-hidden>
Date: Thu Jul 08 2010 - 05:11:47 EEST

On 07/07/10 18:42, James Morris wrote:
> Hi,
>
> Yoshimi 0.58 crashes when loading a padsynth instrument from instrument menu.
>
> The attached instrument fails, as does a simple test instrument with
> default params.

hi James, well that turned out to be more interesting than a boring little
deadlock. With plinky-plank's sample-quality->samplesize of 5, it sets about
prepping an fft plan on 512K, and any way you go about that I figure that's
going to take a finite chunk of time. My terminology use here is no doubt way
off, I don't pretend to understand as much as I'd like on the subject of fft
processing.

The "problem" was that somewhere way back there I switched the planner flag
from FFTW_ESTIMATE to FFTW_MEASURE, and as the docs put it, "FFTW_MEASURE
tells FFTW to find an optimized plan by actually computing several FFTs and
measuring their execution time. Depending on your machine, this can take some
time (often a few seconds)". On my AMD250/4G ram, plinky-plank took all of
1'24" to come to life. And wasn't it worth the wait!

By switching the planner flag back to FFTW_ESTIMATE, things happen a whole lot
quicker. This one is just further evidence that when it comes to the core
synth stuff, Paul (original zyn author) actually knew what he was doing, and
every time I've tried to vary things in that area I've stuffed it up! I think
this little item might also account for comments along the way that instrument
loading in general was taking longer than it used to.

And so there's ...
0.059-pre2 switch fftw planner flag from FFTW_MEASURE back to FFTW_ESTIMATE,
enabling plinky-plank to load in realistic time.

Let me know if it works for you, <http://www.graggrag.com/yoshimi/yoshimi-0.059-pre2.tar.bz2>

cheers, Cal

  
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/listinfo/linux-audio-user
Received on Thu Jul 8 08:15:01 2010

This archive was generated by hypermail 2.1.8 : Thu Jul 08 2010 - 08:15:02 EEST