From: Simon H. <sha...@vi...> - 2006-11-07 08:22:34
|
hi daniel > As for the algorithm I used, I took the knowledge I had gained working > on the occlusion culling for 2.0 and used it to create the algorithm for > switching LODs. It basically takes into account how much screen size an > objects bounding box would take up and also how fast the triangles > degrade within the node. It seems to work pretty well and the objects > that are taking up the most screen real estate are always rendered in > high detail. i took a look at the new code, nice work. in ScreenLOD::render(Action *action), you compute the bounding rectangle of the projected, axis-aligned bounding box. did you try other methods? e.g., i am curious how very long and thin objects perform. unfortunately, i am currently working on 1.x cvs, so i cannot easily try it out. cheers, simon |