[osg-users] Searcing for raytracer implementation based on OSG

Jean-Sébastien Guay jean-sebastien.guay at polymtl.ca
Thu Oct 11 06:51:50 PDT 2007

Hello Adrian,

> i can send the code tommorrow, then you can review it, test it
> and my we can improve a new boosting algorthime for next
> generation of OSG.

That would be great. In which timezome are you? I am in GMT-5, Eastern  
Canada/USA timezone.

> How would look like your timetable, do you
> feel like having some time left to work on the osg integration or
> do you just like finish the master work?

Honestly, I am getting close to finishing my Masters (still maybe one  
or two weeks coding and then I'll finish writing my thesis) because I  
will probably start working soon. Also, since the goal of my Masters  
is not to get something really fast, but to see if I can prove my  
hypotheses, any acceleration of the intersection testing will be an  
added bonus, but not essential.

So what I'm thinking is that I'll test out your changes, see if they  
work well. If I have trouble I'll have to evaluate if I have the time  
to investigate now or just go on without it and work on it after my  
Masters is done.

That being said, I'm very hopeful, because right now it's really  
painfully slow! :-)

> what fps you have at the moment, how many tri-ray intersections
> you have to calculate for a default scene?

I currently only do intersection tests as a preprocess, so the fps  
does not depend on the intersection testing. In one case I have a  
scene where I need to do about 4 million intersection tests as a  
preprocess, and this takes about 200 seconds (a little over 3 minutes,  
or about 20000 intersections per second). Of course, the  
IntersectionVisitor uses the bounding sphere/box hierarchy to avoid  
doing ray/tri tests when they are sure to be negative, so it is not  
20000 ray/tri intersections per second. But that gives you an idea of  
my current situation.

For some upcoming experiments I expect I will have to do much more  
intersections (up to about 100 million) and I will soon integrate some  
intersection testing at runtime (not as a preprocess, but decoupled  
from visualization) so the added speed you speak of will _really_ be  

Thanks in advance, I look forward to trying out your changes

Jean-Sebastien Guay     jean-sebastien.guay at polymtl.ca

This message was sent using IMP, the Internet Messaging Program.

More information about the osg-users mailing list