crazy slow about reflexityai

[FBast]

A large bug with QueueAI coroutine starting for every ai in the game (instead of one coroutine for all ai) were causing a huge drop in fps, now it's totally fine, you can use the benchmark to try it.

[FBast]

Hi laurentopia, in a previous version I used async treatment and queuing stack to optimize reflexions, this way I could mitigate CPU overload with a larger time between each individual reflexion. I have rollback this update for multiple bug encounter, now GC Alloc is at his minimum value and optimization can only be reached (in my pov) with async optimization. I will take a day or two for a large optimization commit, but I can't find any time for now. This project is largely based on reflection so it can be a pain in the ass to optimize correctly. If you want to fork it and try an optimization commit feel free to do. I'm interested by your results.

[FBast]

In fact i also remember using node data cache to prevent fetching reflection data but i rollback it too =/

[laurentopia]

Well done.
Are all the cubes running full AI on each frame?

[FBast]

An AI will run full process when his turn come on the queue, this mean that the more AI you use, the more the time between each of his reflexion is long. This allow to use a lots of AI with no framerate drops.

[laurentopia]

A load balancer, this is nice. Can they be prioritized by proximity to the camera?