On 10.03.2022 18:28, Ichthyostega wrote:
This makes me think into another direction: Maybe just the innards of the
PADnote are too cohesive? We could think of the wavetable interpolation
(which executes the innermost loop) as a distinct component.
...
Am 10.03.22 um 19:51 schrieb Kristian Amlie:
Although I'm not deeply involved in this thought process, I can just mention
that Yoshimi is using several distinct variants of the wavetable
interpolators in ADnote, depending on whether modulation is involved, what
kind it is, and whether the voice is imported into another voice. So doing it
that way fits neatly into the existing pattern. It's chosen by conditionals
and direct calling though, not virtual functions, but the principle is the
same.
That's an relevant observation indeed. Moreover, the PADsynth uses a simplistic
implementation, while the ADDsynth uses a cleverly optimised variant of
basically the same scheme (which was cherry picked from ZynAddSubFX).
A comment there claims that that optimisation gives 15% performance gain;
not sure how this was measured (microbenchmark or benchmark of the whole
computation?), but it might be an interesting idea to use the same optimisation
for PADsynth too. (Won't do that right now, just indicate the possibility).
-- Hermann
_______________________________________________
Yoshimi-devel mailing list
Yoshimi-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/yoshimi-devel
