[Bf-cycles] Interested in opencl

[Sergey Sharybin]

Hi,

There's only rather small portion of Cycles written in pure OpenCL, and
it'll be changed in the future as well. The idea of Cycles kernel is that
the code is shared across all the compute devices and just being compiled
with different compilers.

I'm not sure how much feasible to make pre-GCN cards working correct,
simply because you might easily run into compiler bugs which you can not
avoid. But maybe the split kernel work which is now in master branch will
help you there.

As for mixed CPU+GPU rendering you can do it with OpenCL quite easily, but
it's not really efficient solution. Ideally it need to be implemented as a
Cycles' MultiDevice which uses both GPU and CPU devices. But doing this is
a bit more involved than it might sound.

As for reading about Cycles, best thing to start with will be
http://wiki.blender.org/index.php/Dev:2.6/Source/Render/Cycles
As for reading of path tracing in general, there are lots of information
online, for example PBRT.


On Mon, May 18, 2015 at 1:49 PM, Josh Ring <frag180 at gmail.com> wrote:

> Hi Everyone,
>
> My name is Josh Ring and I'm interested in opencl and cycles, this is
> definitely not the most simple combination. Wondering where to start I have
> experience with c/c++/fortran ultimately I would like to work on the opencl
> portion of the code and get my pre GCN card working and I am interested in
> hybrid cpu-gpu rendering via task parallelism.
>
> I was planning to study ray tracing in general, opencl in general and then
> ray tracing via opencl, does anyone have any recommended reading on these
> topics, anything cycles specific for instance?
>
>
>
> Thanks
>
> Josh
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>


-- 
With best regards, Sergey Sharybin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20150518/acd2b8fd/attachment.htm