Re: [Algorithms] Portal VS BSP
Brought to you by:
vexxed72
From: Alen L. <ale...@cr...> - 2003-12-15 13:24:48
|
> At runtime we recursively cull and clip portals agains eachother It is interesting to know that the method is usable. It was deemed as too much overhead when I read about it a few years ago. The portals need to be convex, though. > Our artists just build sectors of about 20.000 polyons, those sectors > are connected with portals. A portal is a convex planar polygon. When you say 20.000 polygons, you mean that you get average of 20000 rendering polygons per one visibility sector? How complex is the actual sector geometry? Or are you not using any occluders, only portals? I believe, from your comments on portal clipping, that you don't need it. In that case are your artists manually tagging polygons wrt which sector they are in? And how are dynamic objects classified? > If intersection is > empty or zone already processed, stop recursion. When the intersection > isnt we cull each part of a zone using the new frustum and recurse using > the portals of that zone. Are you saying that you actually cull polygons inside a zone, based on clipped frusta? Isn't it too much overhead to even touch each rendering polygon with the CPU? In our tests, even on a fast PC, when polycount gets large, it is better to just throw the entire zone at the GPU and let z-buffer handle it. Alen |