Future of Shiva, lack of future of OpenGTL

[Cyrille Berger Skott]

On Sunday 01 December 2013 11:39:20 Boudewijn Rempt wrote:
> On Saturday 30 November 2013 Nov 09:35:17 Cyrille Berger wrote:
> > Most definitely. I don't have an explicit roadmap yet. I also haven't
> > follow closely enough to know if you already have figured out a way to
> > integrate OpenCL in Krita, but in any case, I would start with
> > integrating the C++ backend first.
> 
> Ack... For opencl, there's a branch by Matus that uses the qt bindings to
> opencl. This isn't ideal because it goes through QImage. We haven't 
done
> anything else yet.

You actually don't need to go through QImage with the qt bindings, you can 
give a "char*". But well anyway, for Krita that is not the biggest challenge, 
the big challenge is to have a full opencl pipeline, it doesn't really make 
sense to move between CPU and GPU. And this transition is difficult, just 
look at how long it takes for the Gimp to move to Gegl... I guess I have it 
easier by starting from scratch.

That said, Shiva could help there, if it has a path to C++ and a path to 
OpenCl, if we can write filters in Shiva, then we can use in C++ until we are 
ready to move to OpenCL. And I guess Krita is in the same situation as my 
application, it needs to be able to run on non-OpenCL devices. The biggest 
challenge is that Shiva is not color space independent friendly, I haven't 
figured out a good way to handle that, I am already struggling a bit with 
RGB and Gray, and both are actually the same color space.

Well anyway, this is far future plan, maybe we will have the opportunity to 
talk about that in Copenhagen.

-- 
Cyrille Berger Skott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kimageshop/attachments/20131201/edc10149/attachment.html>