A question about speed:
what will be faster in a scene with high-polygon-count but over a huge area.
(imagine an outdoor enviroment, a town or so)
will it be good to build a CVA of the area (a big object) or is it faster to cull every vertex against the viewing-frustrum and draw it in immediate-mode?
thanks for help!
Bastian
If you have something like a town, try culling objects like buildings against the viewing frustum. That way you can have each building still rendered with a vertex array and the culling should be cheap too, if you use something like bounding boxes for the buildings. If part of a building is in the viewing frustum, just throw the whole building at the card and let the card handle it. Of course, whatever method you choose will be highly dependent on the app.
Hope that helps.
(1)
Frustum Culling and immediate mode rendering is not fast at all. I’ve used it just to get my culling system to work. Not very pleasing frame rate.
(2)
As ffish says, culling primitives that get shoved into the render by frustum culling is a much faster solution.
(3)
You can also copy vertex indicies into a index_array in your frustum culling algo, and render a small set from an big array. This is what I do right now for rendering an static arbitrary model.
But smaller objects (in size not polygons) will probably benifit from (2). I will implement (2) for all dynamic (moving) and static object, except for one big world object that the whole level is based on, on which I will use (3) or split into sectors.
Just my thoughts on this problem 
Good luck McB
[This message has been edited by julius (edited 09-20-2001).]
[This message has been edited by julius (edited 09-20-2001).]