Re: [Algorithms] How to get 3dvector largest coordinate index?

[Sylvain G. V. <vi...@ii...>]

From: Stefan Sandberg <kef...@gm...>
> By all means, elaborate on what you're doing, if you don't mind... :)

Well I was simply trying to get rid of as many computation as possible
in my just-for-fun home-project of real-time global illumination
approximation. I'm computing that on the CPU, per vertex.

I'm of course using an octree to find out which vertexes "see" each
other (then computing GI is simply computing vertex colors and
propagating them to connected vertexes). The idea behind my previous
question was to split each octree cells depending on the normal of the
vertexes. Actually I'm now using a very simplified way to do so:
>           RoundingOf(1+normal.x) + (normal.z<0.0f ? 3 : 0)

I don't split my octree cells depending on the normal.y for simplicity.
Now I'm down to 1.53 seconds for the Sponza Atrium, and 3.16 seconds for
the Sibenik Cathedral - models from Marko Dabrovic available at:
http://hdri.cgtechniques.com/~sponza/
(these models are not very clean: lots of duplicated vertexes, not
shared edges, holed, inverted faces creating holes...)

My project is still in a very early stage, so the results are not that
impressive yet ; with only one light in the scene:
faked-GI: http://tfpsly.free.fr/Files/SiCath_gi.jpg
Classical diffuse+ambient: http://tfpsly.free.fr/Files/SiCath_ambient.jpg
Only the diffuse: http://tfpsly.free.fr/Files/SiCath_noambient.jpg


> Sylvain G. Vignaud wrote:
> > From: Jon Watte <jw...@gm...>
> >   
> >> Sylvain G. Vignaud wrote:
> >>     
> >>> I didn't need such function before, so I'm not sure if this is
> >>> considered fast or slow. Do you guys have something faster?
> >>>   
> >>>       
> >> What hardware are you running on? Does it have conditional 
> moves? 
> >> Is it 
> >> in-order or out-of-order? Does it have memory pre-fetch?
> >>     
> >
> > PC, current gen. It's for a real time pseuso global illumination
> > attempt. I currently need a 10 second preprocess step to compute how
> > vertexes "see" each other on the Sponza Atrium or the Sibenik 
> > Cathedrals from http://hdri.cgtechniques.com/~sponza/
> > Which means I can have dynamic lights, but my models are to be 
> > static. I'd like to see how much I could speed that up, even if 
> > dynamic meshes seem quite far away.
}