What kind of VSD algos are people using in their 3D engines (if they are making one). I am using portals which I create from a BSP tree. -Icarus

You are using Portals, well, What model files are you using, and how do you know what polys are in which sector/cell, and how do you model them. Im sorry I didnt answer your first question, Im just VERY curious about how your doing it, thanks for your help Icarus!

I am using Half-Life .map files. I have an explanation of how I calculate all of this here:

http://www.gamedev.net/community/forums/topic.asp?topic_id=12994&forum_id=12&Topic_Title=Building+a+PVS+for+a+BSP+tree%3F&forum_title=Graphics+Programming+and+Theory&M=True&S=True

I am sorry that I do not know how to make it a link.

I am having a few problems, and I am not quite satisfied with the output. Even though it is optimized to choose the splitting plane that splits the least number of polygons, the program still splits too many polygons. Ahh well...

-Icarus

Edited by - Icarus on 5/7/00 1:13:44 AM

Hey Icarus,
This may seem really obvious,
but here it is anyway. You do realize, just choosing the polygon that splits the least at each level is not enough. A polygon that may have been split at the last iteration may split less than a polygon that would have been chosen if the polygon in question was never produced.
For example. Let''s say poly A splits B, but C doesn''t. By choosing C, it may leave a set of polygons that everyone splits very many. But if I choose A, B'' may have split less. But then, at the next level, B''''s set may lead to many splits, while if you used C, it would have been better. I came upon this while working on my thesis a year ago, but didn''t really bother looking into it since it was not part of the topic. Same thing went for the PVS on a BSP tree.
You may have been aware of this problem, but if you
did already come up with a solution to this NP complete problem, I would love to hear it. You solution for the portal generation for BSP trees was great!

Phil