[ODE] Ways to contribute brainpower to ODE?(Higher order integrator?)

[Erwin Coumans]

Hi,

For Sphere-Sphere a custom solution is likely to be faster.
GJK is an excellent way to go for combinations that involve Convex Hull, 
Cylinder, Box, Triangle and Cone.

You can check out Bullet GJK performance and stability yourself:
http://www.continuousphysics.com/ftp/pub/test/physics/Bullet-1.3-win32-demo-April2006.zip
It has mouse picking and performance counters ('p') to examine stability and 
performance.

The previous Bullet-ODE had performance issues.
I'm going to provide a new faster GJK-ODE sample integration very soon.

Erwin
http://bullet.sourceforge.net


----- Original Message ----- 
From: "Rodrigo Hernandez" <kwizatz at aeongames.com>
To: <jsinecky at tiscali.cz>
Cc: <ode at q12.org>
Sent: Sunday, April 23, 2006 10:22 AM
Subject: Re: [ODE] Ways to contribute brainpower to ODE?(Higher order 
integrator?)


>
> I cannot say for sure whether you're right or wrong, but your conclusion
> is the same one I arrived at, specialized tests might be less flexible
> than a general one,
> but so far, they're less inestable and take less CPU cycles than at
> least the (general) GJK algorithm.
>
> Jaroslav Sinecky wrote:
>
>> Yes, for convex hull there is already Bullet library implementing GJK
>> and the new implementation by Rodrigo. I tried the Bullet-ODE
>> integration couple months ago, and it worked, but my frame rates
>> dropped deep deep bellow interactive. I don't know if it got any
>> better since (Erwin?)
>> At the end, my impression was that costume collision code for each
>> primitive pair will probably always beat general convex hull
>> algorithms in terms of performance. Or am I wrong?
>>
>> Jaroslav
>
>
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode
>