[Bf-committers] BIH progress update

Martin Poirier theeth at yahoo.com
Tue Mar 25 21:42:47 CET 2008


--- Yves Poissant <ypoissant2 at videotron.ca> wrote:

> There are three factors in a scene characteristics
> that affect how it 
> performs with octree vs BIH.
> 
> The first one is the distribution of the primitives
> in the scene. Some form 
> of statistics could be computed regarding primitive
> distribution per scene 
> to decide which acceleration structure to use. But
> computing those 
> statistics take time and just delays rendering.
> 
> The second factor is the surface area of those
> primitive which must be taken 
> into account and that complicated the statistics
> quite a bit.
> 
> Then the third factor is the orientation and the
> thinness of those 
> primitives. Long thin primitives oriented diagonally
> relative to the scene 
> bounds are hard for BIH.
> 
> So the intelligent way of selecting the right
> acceleration structure involve 
> computing a rather complex set of statistics about
> the scene characteristics 
> and that takes quite some time. So we would just end
> up with even longer 
> render times in the end.

If it comes to down to that, those statistics could be
calculated as a one time pass though. There could be a
function somewhere to "optimize rendering" that would
calculate all that and then pick the best solution
(eventually, it could do other things like reducing
samples when unneeded, reducing buffersize, ...).
You'd run that once and then be done with it, unless
you subsequently modify the scene a lot.

Martin


      ____________________________________________________________________________________
Be a better friend, newshound, and 
know-it-all with Yahoo! Mobile.  Try it now.  http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ


More information about the Bf-committers mailing list