From: Chad A. <ae...@vr...> - 2002-11-18 07:44:53
|
Honestly, I don't think that comes anywhere close to being a performance bottleneck right now. In today's hardware, 10,000 polygons is simply nothing. A GeForce 2 can pump 25 million polys per second. That's 416,666 per frame. According to the profiler, we still spend a very large chunk of time doing collision detection. I bet we would get large speed-ups doing the following two things: 1) distance culling of objects so we don't draw things that are far away 2) a colldet algorithm that isn't O(n^2). we could do some sort of mass culling or at least a sweep test. Honestly, I think we would get a decent speed-up (although last time I ran the profiler, we only spent 20% of our time in the draw() function on *MY* computer) simply replacing OpenSG with our own renderer. OpenSG is rather threaded, and the cost of context switches isn't cheap. In fact, I plan to add support for unthreaded Audiere to see if it helps improve the framerate. Josh Brown wrote: > that our bullet casing has 100 polys in it, I figure when we are running > through the 3 turrets at the begining of the game that there are around > 100 casing sitting on the ground that a wopping 10000 polys that should > only be 600polys (at the most and probably only 300). Anyway this is > defenetially a performance hit that we should have caught earlier. > > J > > > > ------------------------------------------------------- > This sf.net email is sponsored by: To learn the basics of securing > your web site with SSL, click here to get a FREE TRIAL of a Thawte > Server Certificate: http://www.gothawte.com/rd524.html > _______________________________________________ > Isugamedev-midworld mailing list > Isu...@li... > https://lists.sourceforge.net/lists/listinfo/isugamedev-midworld |