Hi, I’m studying collisions recently and doing some performance test
Here’s the setup:
A boulder mesh with 3000 triangles
With simple collision of 16 convex hulls
Which only blocks pawn
I place dozens of it in the same place in a level
Toggle stat.Collision and watch the SceneQueryTotal value
Then something strange happened when my pawn stands on the pile of meshes
With simple collision the query time is larger than that using complex collision
It contradicts with my understanding of collisions:
I assume that it first checks if the collision is possible using bounding box, if possible it will test for all triangle planes.
So if using simple collision it probably don’t need to test all 16 convex hulls, even if that’s the case, the number of total triangles is not likely to exceed 3000, where 3000 is the number of triangles complex collisions need to test.
The performance of simple collision must be better in most cases, while the result is on the opposite side when my pawn walks around.
I’m confused with the results, am I getting something wrong?
Thanks.