Hi everyone, I wanna know the thoughts of people who have implemented or have experience with physics sims and/or collision detection routines, on the retention and data representation part of the equation.
I've recently poked PhysX to see how it can be integrated to my 3D game, but immediately realized that i absolutely hate the way it wants me to keep all state in sync, and it gets ugly very quickly, not to mention all the copying and conversions that should happen every frame. All because it keeps an internal representation called "Actors" to make everything magically work after simulate(); . I know this gives the API more control over its internal simulation and optimization, but it becomes trouble for the app that's using it. As if the API was made with the expectation that the user only loads a scene once and never wants to inject changes into it...
Now I've decided to make my own collision detection/physics without use of any retention, and just use the entity state as the input to test functions.
But Considering that i haven't implemented any physics sim yet, i can't foresee the requirements and challenges, so i don't know.. even retention might become necessary. and my game doesn't depend much on physics nor does it use any substantial number of rigid bodies. for now i'm just thinking about simple collision detection.
Any thoughts?