Re: BVH instances support : approaches back

(L) [2013/09/11] [ost by Serendipity] [Re: BVH instances support : approaches] Wayback!

I am not sure what your target audience will be by in my opinion you will most likely need version 1 anyways if you want to support dynamic scenes with many triangles. Every rebuild algorithm I know is still to slow to handle 50+ MTriangles at 30fps or so. I also don´t understand why spatial splits should not work. You can still use spatial splits for each object. From my experience the performance does not suffer very much when using a 2 level hierarchy unless you have many overlapping objects.

(L) [2013/09/11] [ost by spectral] [Re: BVH instances support : approaches] Wayback!

You're right...

I don't care about rebuild time for now... I'm only interested by offline rendering.

Maybe I was not clear, but spatial split is not possible for the TOP level BVH because it only contains instances, but still possible for sub-SBVH ! It is just that in this case I cannot compute the spatial-split on all the primitives (because I have several sub-trees).

Remarks: yes, overlapping objects is a problem in such case... not sure that there is a way to optimize this !

(L) [2013/09/13] [ost by graphicsMan] [Re: BVH instances support : approaches] Wayback!

BTW, if there are students interested in publishing a paper about how to solve this problem, I would be interested in helping out.  I've been interested in this for a while, but I simply don't have the time to do a full implementation of my ideas.  If you're interested in pursuing this, and would like some help with concepts, paper writing, and perhaps a small amount of coding, send me a private message.