This topic is now archived and is closed to further replies.


AABB trees splitting

Recommended Posts

I was just wondering how everyone does there tree splitting, if most use a 3rd party source, or just ideas on this? For instance, if you use a common method like splitting based on the tris midpoints average but you have interesting ways to return colliders... for example like if you stored each tris array position per node you could devise an optimal algorithm taking into account depth, number of returning tris, and time to calculate final per tri test compared to time for an object AABB box... I have done alot of online searching too, so if you are using Opcode, Rapid, or some other one I am familiar with its workings, just wondering if someone knows of a original way. Doesn''t need to be better, just different. Also if anyone is interested I wrote an AABB Octree from stratch that is pretty fast and accurate enough for collisions and reasonable enough for frustrum (I say reasonable since for frustrum you would probably just use a Vertex type test and not AABB). AABB seems to create 10% or so more vertices in the leafs which is expected since it is picking up exact triangles but for a frustrum you don''t need to be that accurate and wouldn''t need to deep of a tree.

Share this post

Link to post
Share on other sites