(Was: Re: koffice/krita/image/recorder)

[Cyrille Berger]

On Sunday 24 January 2010, Cyrille Berger wrote:
> On Sunday 24 January 2010, Boudewijn Rempt wrote:
> > On Sun, 24 Jan 2010, Cyrille Berger wrote:
> > > Sure 2.2 isn't going to be the time to freeze pigment, but I am all for
> > > doing the change we want the sooner. As you have noticed, I have
> > > already done quiet a few API adjustement over the last two weeks. And I
> > > intend to keep going on, at least until end of february.
> >
> > Then if you think the koffice release manager won't object, go ahead, by
> >  all means :-).
> 
> hehe. But the question is wether we really want to ?
> 
On the performance question, I wrote a benchmark in the pigment directory. In 
attachement comes results I had.

alpha / setAlpha = 8bit
alpha2 / setAlph2 = qreal

the setAlpha is a single function call while for "setAlpha (Individually)" the 
function is called once per pixel.

No suprise, alpha is faster for 8bits than alpha2, and alpha2 for float 
colorspace. setAlpha2 when call once has the same speed, again not surprising 
since the opacity is converted to the native depth only once.

Now the question would be to study our use of those function, and 
alternatively we can keep both functions.
-- 
Cyrille Berger