[Bf-committers] BIH progress update
Joe Eagar
joeedh at gmail.com
Thu Mar 20 15:31:00 CET 2008
Code specific to only 1 architecture isn't allowed. However, I think if
you wrote a generic SIMD macro library and used *it* then it'd be ok.
At least that was what I was planning on doing back when I was
considering using SSE for something (using SSE by itself had already
been vetoed at that point, but I figured a generic macro library could
support Altivec too, so it wouldn't really be architecture-specific
anymore).
BTW, do you have a link to that forum post? It sounds cool.
Joe
Yves Poissant wrote:
> This is a head-up to give a status report about the BIH implementation I'm
> working on.
>
> I just spent the last 5 days working on BIH. Not much progress
> unfortunately. On some file, I get spectacular improvement, up to 4 times
> faster render times than octree. But on some other files, I can get
> humongous render times. On average, there is very little gain. I tried all
> the optimizing tricks I know, including caching the quad data, but that
> gained very little too. I rewrote the tree building algorithm to allow me to
> test different partitioning heuristics but I have not started on that yet.
> I've added code to cut out empty space but that deepens the tree and
> lengthens the total render time. As I added code to cut off empty space, the
> partitioning process looked more and more like a defered BVH so I'm begining
> to think that I should try a SAH BVH just to get a comparison point. I
> haven't touched on the ray-triangle intersection aspect yet because the
> partitioning complexity is way too high right now and that needs to be
> optimized first. I need get data out of the partitioning and traversal for
> the scenes that takes huge render times to figure what makes those scenes
> behave that bad.
>
> I thought of giving this head-up because I will now have to return to my
> regular work for a while but I will be back on the BIH later. I'm sure the
> back burner will keep the issue warm in my head though.
>
> What is the policy about using SSE ? Someone posted an SSE function on ompf
> forum that computes intersections between one ray and 4 triangles in
> parallel and at some point I'd like to try that. Since quads are 2 triangles
> and I currently store 2 quads per leaves, that would fit very well. SSE
> could also be used efficiently when doing bounding boxes operations.
>
> Yves
>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>
>
More information about the Bf-committers
mailing list