[Bf-cycles] BVH White Papers

[Brecht Van Lommel]

Hi,

Our implementation is in fact based on the SBVH code from NVidia, same
as the first paper. We now have some code from Intel too to optimize
the building and traversal code further, but the overall idea is still
there. The MSBVH is an interesting optimization to improve the BVH for
motion blur. That can be looked at once deformation motion blur is
working.

No rewrite is needed, once you are more familiar with the BVH
implementation you'll see how these algorithms can be added to the
existing code. First focus on implementing a "moving triangle"
primitive in the existing BVH, and once you have deformation motion
blur working with that you can look at deeper changes to the BVH if
needed.

Brecht.

On Wed, Jun 19, 2013 at 4:47 PM, Gavin Howard <gavin.d.howard at gmail.com> wrote:
> Hello all,
>
> I have been working on finding the best way to implement the BVH for
> motion blur in Cycles. I have found two white papers that may have a
> good answer. The first is just a new way to do BVH's in general. The
> second is a way to extend the first for motion blur. I have uploaded
> the pdf's to Dropbox. They are at
> https://www.dropbox.com/s/y61dsnzhqpw77uz/sbvh.pdf and
> https://www.dropbox.com/s/4e4q4sqk99popuy/msbvh.pdf.
>
> The important thing about doing it this way is that I would have to
> rewrite Cycles' entire BVH code. Brecht, Stuart, Ton, I would love
> your thoughts. The advantages I see are the less intersection tests
> and less traversal, as well as having a way to extend it for motion
> blur fairly easily. The disadvantage is that I would have to rewrite
> the BVH building AND traversal code for Cycles.
>
> This is just an idea, by the way. I don't really expect that this is
> what devs want. If it's not, I will find another way.
>
> God Bless,
> Gavin Howard
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles