PD patch complexity and Lich efficiency

Hello, hope you’re doing well
I’m very excited about the opportunity to play generative patches by the means of this brand new technology. So I have one main question regarding its capability within the Lich module.
I listened to very big patches made out of Pure Data over the years, some were rather “heavy” and drag into mess the cpu… Since I imagine that within the module the patches will be more subject to interplay with the outer control voltage world, I wonder how far one can go with the “complexity” of a patch to be loaded and performed seamlessly onto the Lich module. Could you please spend few words concerning eventual experienced limits, especially compared to a given PC/Mac configuration?

Hi @freemod,

If you have a patch that is too slow for desktop, there’s no way a 180MHz microcontroller would be able to run it. There are a few other considerations:

  1. OWL doesn’t run the same PuraData that you run on desktop - this is not possible, because it would require an actual OS. It uses a compiler called “Heavy” that generates C++ code from PD that getss compiled to run on microcontroller.

  2. Not every object is available when compiled on Heavy, you can find more info in our PD docs

  3. Some parts of generated code would be using optimized maths calls, this improves performance (sometimes considerably). This may also cause some differences in patch output, but those optimization can be disabled in compiled code.

I don’t use PD myself, but I’ve seen some impressive stuff done it in for OWL, i.e. you can check out generative Fascination Machine patches in patch library.

You can try your own patches in web patcher, once you’ve got something that works you can ask here for someone to test it. Then you’ll get feedback about its performance/usability.

thank you very much for having fully clarified