Krita's rendering model

[Boudewijn Rempt]

On Thursday 15 September 2005 14:42, Bart Coppens wrote:
> On Thursday 15 September 2005 14:11, Boudewijn Rempt wrote:
> > So if there are two views, or three views, we composite two or three
> > times.
>
> Yes that's a problem indeed.
>
> > 1.
> > Image creates QImage from the dirty rect ->
> > Image emits a signal with the QImage as a parameter and the dirty rect
> > Clients connected to this signal take the QImage and paint it onto their
> > canvas
>
> Good as it sounds, this has a problem. As I understand, each view has an
> exposure setting for the float colorspaces. Compositing the layer once and
> then emitting it would effectively kill this parameter to be differentiated
> among different views.
>
> > 2.
> > Problem: this cached QImage can become pretty big, of course, and that
> > would take an unpleasant amount of memory.
>
> Again, the exposure is not taken into account. And of course, this gigantic
> waste of memory is not so fun.
>
> But maybe you could cache the KisImage's renderToProjection result between
> different calls to renderPainter that are made after the same update of the
> image?

Good idea! A sort of expiring cache for rects with the exposure parameter 
added, that should work very well.
-- 
Boudewijn Rempt 
http://www.valdyas.org/fading/index.cgi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://mail.kde.org/pipermail/kimageshop/attachments/20050915/072059a9/attachment.pgp