[Bf-cycles] Cycles kernel split patch

[Marc Dion]

bug handling and continued effort... thank you very much :)

(i)   Where should the bug reports go?

(ii)  It appears as if even the most basic scene takes up a large amount of
GPU memory (simple model, no textures) .

There do not appear to be any GPU optimization specialists currently active
in Blender development.

Would it be possible for you to shift your efforts towards achieving
maximum compatibility(for device's/OS's) vs taking full advantage of the
feature set that some flagship cards offer so that even reference level
hardware with limited threads and only 256-512MB of RAM can also benefit
from this?

Well anyways, thanks again.  This is some of the best news Blender has had
in quite some time and the buzz is already well underway.  You've made a
lot of people very happy with this work you've done.

PS.  Jens V. seems to think that CPU rendering is not setting number of
threads correctly. Too many or something like that.










On Wed, Mar 25, 2015 at 7:03 AM, Kyriazis, George <George.Kyriazis at amd.com>
wrote:

>  Ton,
>
>  thank you so much for the kind words.
>
>  bf-cycles,
>
>  We have attempted to split the OpenCL kernel with the utmost goal (of
> course) to get better performance on both discrete GPUs and APUs
> (integrated).  There was been a lot of work spread out over different kinds
> of optimization types in order to get the performance improvement.  This
> has been detailed in the doc that Ton provided below.
>
>  AMD is definitely planning on continuing this work (fixing bugs,
> providing additional performance improvements, etc.), so please do treat
> this as an ongoing effort.
>
>  I'd like to express my thanks to Ton and Martijn for being supportive of
> our work, and of course to our engineering team for the hard work to push
> this out.  They'll definitely continue to be active for ongoing support.
>
>  I'd like to remind everyone that feedback, bug reports, etc. are always
> welcome.
>
>  Thank you,
>
>  George
>
>  On Mar 25, 2015, at 7:58 AM, Jonathan Williamson wrote:
>
> This is really great. Thanks AMD!
>
> On Wednesday, March 25, 2015, Ton Roosendaal <ton at blender.org> wrote:
>
>> Hi all,
>>
>> A couple of developers in AMD have been working for many months with
>> trying to get smaller OpenCL kernels for Cycles. Yesterday they've posted
>> their patch for review:
>>
>> https://developer.blender.org/D1200
>>
>> Doc:
>>
>> https://docs.google.com/document/d/1LuXW-CV-sVJkQaEGZlMJ86jZ8FmoPfecaMdR-oiWbUY/edit
>>
>> Whatever the outcome is, I'd like to express my thanks to AMD and George
>> Kyriazis for their efforts. A corporate contribution to Blender of this
>> scale is unique in our history. I hope it will be an inspiration for others
>> to consider to help us out as stakeholders as well!
>>
>> Laters,
>>
>> -Ton-
>>
>> --------------------------------------------------------
>> Ton Roosendaal  -  ton at blender.org   -   www.blender.org
>> Chairman Blender Foundation - Producer Blender Institute
>> Entrepotdok 57A  -  1018AD Amsterdam  -  The Netherlands
>>
>>
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>
>
> --
> Jonathan Williamson
> http://cgcookie.com
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20150325/45e490ef/attachment.htm