The first couple points there are (I assume) trivial, as all I’m doing is storing a single number, and sending it out 256 samples later. So the main bit I was trying to improve in using framelib over vanilla Max is minimizing “the slop of sending it back to the scheduler”. The vanilla Max has the same problem in that it goes from sample accurate triggering to spitting out a number in control rate. So I guess it’s the difference between edge~ + snapshot~ vs fl.tomax~.
At this point you probably need to stay in one domain for your whole processing chain then to avoid any slop by conversion and only inducing delay caused by sample collection. Otherwise you are always going to be waiting an indeterminate amount of time for the CPU to complete its instructions.
1 Like
Oops, I didn’t see that Alex had already mentioned this above!