[ODE] Question about render pipeline culling and ODE Quadtrees.
Jon Watte (ODE)
hplus-ode at mindcontrol.org
Thu Sep 13 16:26:17 MST 2007
It is very common to have totally different spatial data structures (and
even vertex data) for physics vs graphics purposes. In fact, AI often
has its own, third data structure. This is so, because the needs of the
systems, once you optimize them, are very different.
If you want to use ODE spaces for view frustum culling, you can either
implement a Frustum primitive (geom), or you can use a best-fit
geometry, such as a bounding sphere, to do a first-level cull, and then
do the actual frustum cull against the returned set. However, frustum
culling is usually so fast that optimizing it seldom helps frame rates
at all. If you have the number of objects where this starts becoming a
problem (> 10,000), you have other problems, too, such as the batching
overhead when submitting to the graphics card, and the solution is to
aggregate geometries into bigger sets.
Cheers,
/ h+
Jeremy Roberts wrote:
> Hello fellow ODE users,
> I have a question about render pipeline culling and ODE quadtrees.
> I understand that ODE uses quadtrees for quick collision/physics
> culling. In my 3D FPS game engine however, models/objects also need
> to be culled during the rendering pipeline to determine which are
> potentially visible. This is typically done also using an octree or
> quadtree and frustum culling.
> My question is, is there a logical way to use the quadtree that
> ODE uses for collision/physics culling for graphic/object culling in
> the rendering pipeline as well? I.e, frustum cull the units of the
> ODE quadtree to determine which are visible in the view frustum each
> frame?
> My concern is that it seems silly to have to have an
> entire quadtree for collision/physics culling and to have to have
> another entire quadtree for object culling during the rendering
> pipeline if there is a way to use the ODE quadtree for both.
> Thank you for your help.
> Jeremy (grill8)
> [Video Game Computer Science]
> DigiPen Institute of Technology
>
> - Make the world what it could be, not what it is.
> ------------------------------------------------------------------------
>
> _______________________________________________
> ODE mailing list
> ODE at ode.org
> http://ode.org/mailman/listinfo/ode
More information about the ODE
mailing list