Fluid.bufnmf~ "Machine Learning Modulation"

Hello everyone. Getting pin-wheel versions of the non-realtime objects is a matter of running them in their own thread, so that Max’s main thread is no longer blocked. It has to be done in a way that’s portable to SC and PD also.

It will happen, but has to be approached with care: hastily written concurrent code is almost certain to contain bugs and weirdnesses that are much more annoying for you all than the pin wheel, and we don’t want that. The refactoring work that I’m doing at the moment has the eventual need for threading present as a consideration, but first the code has to be brought back functionally to where it was (i.e. what you’ve already got!)

Speed issues are a different matter. We’re starting to profile bits of the code, but – again – it doesn’t benefit any of us to be over-hasty about introducing optimisations. There’s a lot of quite different algorithms running back there, and it would be foolish to try and over generalise what optmisations might be useful.

3 Likes