[ODE] Huge world vs precision problems

[Charlls Quarra]

the general solution of this sort of problems is to keep the geometrical relationships as abstracts as possible and evaluate only when you have to, for example, to send coords for 3d objects. The problem is that ODE is a numerical integrator, so lazy evaluation is not feasible under usual circumstances. A long term solution would be to do a specialized ODE version with support of arbitrary precision numbers, i think there are a couple arbitrary precision libraries online. But then you would have to recast all computations in the LCP into these structs, and convert to 32bit floats only before the rendering stage. I have no idea how the performance would be affected, though

Anders Olofsson <anders.olofsson at biologigrand.ac> escribió:  I tried the double precision version of ode but the problem was still 
there, so.. I just made a little graphic test (OpenGL), 2 rotating 
triangles one at origin and one at the far edge of the world. Ofocuse the 
one at the far edge wobbled, kind of obvious. Then I tried specify the 
triangles using doubles, no difference, still wobbling!. Strange and 
depressing. So the problem isn't just in ODE, duh.

What about scaling everything down to the size wich can be "handled" by 
floats?. Any other drawbacks than the agony? ;)

Thanks everyone.

_______________________________________________
ODE mailing list
ODE at q12.org
http://q12.org/mailman/listinfo/ode


		
---------------------------------
 1GB gratis, Antivirus y Antispam
 Correo Yahoo!, el mejor correo web del mundo
 Abrí tu cuenta aquí
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://q12.org/pipermail/ode/attachments/20060310/97576f57/attachment.htm