Skip to main content
gamedev.net
🔒 Locked

AABB Trees???

Started by gahre Nov 25, 2001 at 3:54 PM 4 replies 700+ views
Original Post
gahre
gahre
Hi! I''m currently using an octree to do visibity checks and collision detection. But now my framerate decreases with the number of enemies (just models moving aroung with triangle-ray intersection testing). Thus I thought that maybe my collision detection routine is too slow, so I thought about using AABB Trees. BUT will it be faster? If so - are there any tutorials or sample implementations on AABB trees on the net? Gahre
Renus
Renus
It''s unlikely that the octree is the slow thing. AABB means nothing more than a box, it''s not a method.

First of all I suggest you to implement a profiler, a piece of code that measures execution time in cpu cycles. This takes some lines of code, but you will immediately see where you loose performance. You should only spent your holy time on code that really need to be addressed next.

- thomas
thomas
phueppl1
phueppl1
But then there are things that just take their time :o) the trick is to know what can be sped up *g*
cya,
Phil

Visit Rarebyte!
and no!, there are NO kangaroos in Austria (I got this questions a few times over in the states )
Visit Rarebyte! and no!, there are NO kangaroos in Austria (I got this question a few times over in the states ;) )
TerranFury
TerranFury
You should probably be able to get away with treating your models as being AABBs and not doing further ray-triangle collision detection.
gahre
gahre
Well, acutally I''m treating my models (objects and characters) as spheres, but I think it''s a bad idea to treat my terrain/world as an AABB
But this profiler thingie is a good idea. Any comments on implementing it?
TerranFury
TerranFury
Spheres are even better!

As for profiling, check out Intel's VTune. That'll do everything for you, and you won't have to code a thing!

Edited by - TerranFury on November 30, 2001 6:46:16 PM

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.

Sign in to reply to this topic.