Change of algorithms for KisPainter::bitBlt

[Boudewijn Rempt]

On Wednesday 28 April 2004 17:25, Cyrille Berger wrote:
> Hello,
>
> I was trying to compare two algorithms for KisPainter::bitBlt :
>  - the older one ( I will call this one OldAlgo)
>  - a new one using KisIterators ( I will call this one IterratorAlgo)
>
> I tried to compare them using valgrind (valgrind --skin=calltree krita
> ~/test.kra and I quit immediately). The two algorithms seems to have the
> same cost when I look at the global cost of krita. But I get some weird
> results : - in the call tree of OldAlgo I don't see any call to
> KisTileMgr::tile - in the call tree of IterratorAlgo I don't see any call
> to
> QPixmap::convertFromImage
> and so on. And I don't understand why.
> Does someone know what are the <cycle 1> and <cycle 4> functions ?

If it's the commented out and if-deffed out code in kis_painter -- I guess 
because the iterator code doesn't use the color strategy. It really 
necessary, by the way to keep all actual pixel computation in the colour 
strategies, because that's the only way to stay color model independent. The 
core of Krita is for accessing pixels, moving over pixels, deleting pixels -- 
the colour models is where pixel values are changed and where we can should 
put computations.

-- 
Boudewijn Rempt | http://www.valdyas.org/fading/index.cgi