[Bf-cycles] raPT, highly efficient BVH kernels

[Brecht Van Lommel]

It's a paper about BVH construction with spatial splits, not
intersection kernels. Of course faster BVH construction would be nice.
I don't think we'll do that on the GPU anytime soon though, as that
paper shows there is still a lot of room to improve BVH build
performance on the CPU without having to deal with the complexity of
doing it on the GPU.

On Sat, Jun 11, 2016 at 8:59 PM, Mohamed Sakr <3dsakr at gmail.com> wrote:
> Hi,
>
> found this thread, http://ompf2.com/viewtopic.php?f=3&t=2098&p=5507
> it states that it outperforms embree, + the margin between both goes farther
> with more threads!
> it is also vectorized (AVX), so this can be implemented in CUDA too I guess
> with a huge boost in performance.
>
> @Sergey and @Brecht, what do you think?
>
> cheers,
> Mohamed Sakr
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> https://lists.blender.org/mailman/listinfo/bf-cycles
>