From: Kristian A. <kri...@am...> - 2012-10-15 20:14:38
|
On 10/15/12 21:33, Will J Godfrey wrote: > On Mon, 15 Oct 2012 14:23:14 -0400 > Mark McCurry<mar...@gm...> wrote: > >>> (I'm guessing it's done in a RT thread now? Since there are clicks I >>> mean...) >> That is how it is done in zyn, at the moment, but this is a known >> issue that will be changed in the nearish future. > > I was under the impression that Cal had taken it out of the RT thread - could > of course be wrong there. With a mega-patch you can certainly see quite a long > delay in updating the UI - for a start the bargraph freezes (so at least that > part must be outside RT), but there is no interruption in the sound. > >>> Is memory really a problem - several 10s of megs is easily goes unoticed >>> within most setups of the last 10 years or so? >> Yes, imagine a bank of all padsynth instruments. > > In discussion we also quickly came to the conclusion that holding even one bank > in memory was a non-starter. However, we were mulling over the possibility of > double buffering so the whole voice modification thing could be done outside the > RT thread then just (in theory) a few pointers swapped at the right moment. Sounds like a good solution. > Cal > ran out of time though :( Hopefully we can finish it for him! > The other half of the equation is that both the DB and my suggestion provide a > seamless, and backward compatible way to get the patches in a form that is > easily exposed to MIDI control. Hmm, is this discussion archived? I might like to learn a bit more about the details. -- Kristian |