[ODE] Huge world vs precision problems

[Jon Watte (ODE)]

Megan Fox wrote:

> The trick works fine for rendering, definately, but to pull that off
> in ODE you're effectively calling setpos on everything before every
> phys world update.  Isn't that a "Bad Idea" (I thought setpos tended
> to infere with the solver)?

I was thinking that you use ODE with doubles for the regular integration 
and "easy" primitive testing, and only move geoms around when testing 
against mesh primitives. In fact, you could keep one of each of sphere, 
box, etc in your code, not in any space, to set position on, and one 
instance of each mesh that's centered at the origin. Inside the 
nearcallback, if testing against a mesh, move/rotate the test primitive, 
and collide it against the mesh instance centered on the origin.

Cheers,

			/ h+