Looking for performance guidelines for my use case

Show what you made with Bullet Physics SDK: Games, Demos, Integrations with a graphics engine, modeler or any other application
Post Reply
toolforger
Posts: 6
Joined: Sat Apr 28, 2012 10:11 am

Looking for performance guidelines for my use case

Post by toolforger »

Hi all,

after having read up on AABB trees and all the nice things that Bullet does, I'm wondering whether what I have in mind is insane or just what Bullet was built for.

The use case is a space simulation (for a game, so realism can be traded off). Stars being orbited by planets being orbited by moons being orbited by spacecraft.
Very few collisions. Say, one collision pair in every 1000 objects.
Everything is in motion at all times.
Very little actual physics - no gravity (that's handled by fiat), no friction, no joints, just impulse transfers between rigid bodies if something is actually colliding.

My question then is:
To what kind of object count can I expect to have Bullet work well with such a scenario? Should I plan for up to 1,000 objects, 10,000 objects, 100,000 objects, or 1,000,000 objects per physics space?
pico
Posts: 229
Joined: Sun Sep 30, 2007 7:58 am

Re: Looking for performance guidelines for my use case

Post by pico »

Hi,

one important factor is that Bullets CPU usage grows linearly with the number of collision bodies, even if they are asleep or static.
On low-end handheld gaming machines you can so easily spend a few milliseconds when you just have a few hundred static objects without any collision response at all.
On PC you can have thousands of sleeping/static objects without much problems.
Post Reply