Hello Karl,
This is very, very fast, <1sec for 1.000 cows.
But some nodes could be outside the selectionBoundingBox, because the box of
the node has an intersection with the SelectionBoundingBox and not the node
itself.
Yes, that's a classic optimization. I haven't looked at the code (that's
the next step) but PolytopeIntersector probably has a broad-phase and
narrow-phase, where broad-phase tests only bounding volumes to get a
"potential intersecting set" and then narrow-phase tests the actual
geometry against the polytope. The broad-phase would use the hierarchy
as well since the scene graph has bounding spheres at the node level and
bounding boxes at the drawable level, so you can eliminate
non-intersecting branches of the graph quickly. It's the narrow-phase
that takes a long time.
I'm not sure if it's done that way, but a potential optimization here
would be to specify to the intersector whether you want a
test-intersection or a find-intersection (the former is just interested
in whether something intersects, not necessarily what - useful for
shadow tests in raytracing; the latter finds everything that intersects
and is therefore longer to compute), and whether you're interested in
individual polygons, drawables or nodes.
My hypothesis is that the PolytopeIntersector wants to return all the
actual polygons which were inside/intersecting the polytope, but in our
case, we just want to know which nodes should be selected. So in that
case, as soon as the intersector finds a single polygon in a node that
intersects, it accepts the node and skips to the next one. I think that
would speed up our situation greatly.
Anyways, as I said I'll start investigating this in practice soon, for
now it's just theory. If you want to dive into the code of the
PolytopeIntersector and see what you can find out, feel free, and report
your findings here please :-)
J-S
--
______________________________________________________
Jean-Sebastien Guay [EMAIL PROTECTED]
http://www.cm-labs.com/
http://whitestar02.webhop.org/
_______________________________________________
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
