branches/koffice/1.6/koffice/krita

[Cyrille Berger]

> * extent() and exactBounds() vs rawData() -- we have discussed this in the
> past. It's not in 1.6, but in trunk we separated the iterators into const
> and non-const interators. This should, in theory, make it possible to set
> the exact extent of an image (by assuming that a non-const iterator writes
> over the whole area it is created on). Not sure whether we did that,
> though. I think not...
No we didn't. The const/non-const things was mostly to enforce that we can't 
write on the source paint device. Now, we probably can use that for making 
optimization :)

> I'm also wondering whether we should do a read-ahead 
> for the iterators, getting the the data out of the tiles ahead of time.
Yes I think it can be a good idea, to pull more than one tile. And maybe if 
it's done in a background thread, we could avoid having krita blocked while 
the tile manager is reading on the hard drive.

> * chunked rendering for filters. I should checkout 1.6 again to see about
> it -- I'm still not sure that putting the chunking inside the filters
> doesn't make life more complicated for filter writers, but I'll check 1.6
> code before I continue with my threaded applicator class in trunk.

isn't the chunking done outside in this patch ?

-- 
Cyrille Berger