some iterator benchmarks
boud at valdyas.org
Thu May 30 08:22:54 UTC 2013
On Thursday 30 May 2013 May 10:12:12 Inge Wallin wrote:
> Assymetric tiles is something I thought of too. It doesn't need to be
> width=2*height either. There can be any factor, and the ultimate in
> assymetric is using x scanlines of the full image instead of rectrangular
Yes... That is something other apps actually do. Photoshop can make either tiles or scanlines available, judging from their old sdk. Cinepaint was going to be ported to scanlines, and nuke uses scanlines for everything.
For painting, though, that is perhaps the worst strategy from a memory point of view.
> And re the undo buffer, does it have to be aligned tiles in there? why not
> make it a quadtree of pixels with varying tile sizes?
Yes, that is something I was thinking about. We could come up with a nicely complicated scheme where we'd coalesce adjacent tiles for every layer automatically, store undo in optimized tile sizes and so on. But before hacking on this, I want us to try an store the image data on the gpu first.
Btw, blender had a very nice comparison for tilesizes, cpu and gpu for the opencl-based backend: http://www.blenderguru.com/4-easy-ways-to-speed-up-cycles/
http://www.valdyas.org, http://www.krita.org, http://www.boudewijnrempt.nl
More information about the kimageshop