Stupid rendering question

You could recursively check every node. Then if it''s visible go down that node, if not don''t go down that node. Then if you''re at a certain depth, or reached the end, add element(s) to the visible parts buffer.

Yes, but my problem is figuring out the order in which the polys are rendered. Do I have to recalculate every polygon''s position relative to the camera? That sounds too expensive...

Thanks again
-OberonZ

You need to copy the elements that get rendered from your tree to a simple "rendering array".
You could make this array a sorted linked list or binary tree, where each element gets sorted into the list/tree when added.

The extra computations you''ll need then are:
- recalculate every polygon''s position relative to the camera
- sorted add, into list or tree

Especially the sorting can take it''s time.

One thing I can tell you when you''re implementing this new linked list / binary tree, is to NOT allocate the memory every time. It''s much faster to use a uge buffer, and do your own "allocation".

But I haven''t coded an octree myself, so this might not be the best way to do it. Maybe some else can explain a faster algo...