collision on faces

[Guest]

I noticed that still the simulation is very dependent on the number of faces in static objects. When subdividing the ram in the demo once, the framerate drops to only 3 frames when the crate and ball reach the pile. Is this going to be optimised?

[Guest]

it will be optimized later indeed. You can either subdivide the object into seperate pieces for now.

the optimization at the moment it:

aabb check per object -> aabb check per triangle -> gjk narrowphase
the aabb check per triangle is still very slow, and perhaps there is some debugging flag which makes the aabb infinitely large

the addition which will improve things will be a kdop tree:

aabb check per object -> kdoptree -> gjk narrowphase