[LAD] LADSPA dilemma

[Paul Coccoli]

On 6/15/07, Fons Adriaensen <fons at kokkinizita.net> wrote:
> On Fri, Jun 15, 2007 at 05:48:13PM -0500, Jan Depner wrote:
>
> >     I tend to agree with Fons on this one but I just want to thank Tim
> > for finding the problem.  I kept running into that one a week or so ago
> > and I thought I was losing my mind.  It kept moving on me.  No wonder
> > since it was using random stuff in memory.
>
> None of my plugins will use any random values, but they will fail
> in the way Tim has illustrated earlier in this thread if used in
> non-RT mode, e.g. when a single run() call is used to process an
> entire file. All those that do interpolate will start processing
> with the defaults set by activate(), and some of them will spread
> the interpolation over the entire first run() call. This will be
> fixed using the 'this is the first run()' flag.
>
> --
> FA
>
> Follie! Follie! Delirio vano è questo !
>

Is it an option to interpolate over a fixed interval regardless of
block size (meaning the nframes parameter to the run() call)?  Or use
the minimum of some small number of samples and nframes.

I implemented parameter smoothing over the entire run() call too when
learning LADSPA.  It seemed somewhat naive, but apparently others did
the same...