Sub-pixel positioning

[Adrian Page]

On Wednesday 17 March 2004 1:04 am, Roger Larsson wrote:
> On Wednesday 17 March 2004 00.17, Adrian Page wrote:
> > On Tuesday 16 March 2004 8:02 pm, Roger Larsson wrote:
> > > > Using valgrind to profile KisToolBrush::mouseMove() with a pipe
> > > > brush, about 80-90% of the time is spent in computeDab(), and around
> > > > 10% in compositeCopy() and compositeOver(). The compositing code also
> > > > has plenty of room for improvement, part of which would be using
> > > > premultiplied alpha for the RGBA buffers. We can also replace
> > > > division by shifts and optimise for each channel depth we support.
> > >
> > > Yea but how much time is spent in X because of what Krita does?
> > > (I had 60%...)
> > >
> > > Please try with my patch (for brushes).
> >
> > I tried your patch, but on my system the lag it introduces makes it feel
> > worse. Perhaps that's why I don't get as much time spent in X when
> > profiling, as for me, X is less of a bottleneck?
>
> Now I remember, I did see less of the problem after I upgraded memory
> (and did remove some of my old stuff)
>
> But you can try with another kis_notify_limit like 10
> (then it will be updated at at least 100 Hz - about what your monitor does)
>
> [make it global - then you can play around with different values in the
> debugger]
>
> How does your system look like?
> * CPU
> * Memory
> * Videoboard (and X server)
> * Painting device
>
> Mine is: Pentium3 933MHz, 512MB, MX 440SE (Nvidia / XFree depending on),
>  Wacom Intous 2 (hmm. it has more pressure levels than most so I guess I
> might be getting bigger patterns)

Mine is Pentium 4 2.4GHz, 1G, GeForce 2 GTS, XFree 4.4.0. That profiling was 
done using the mouse for input but I've just got an Intuos2 too, so once I've 
got that working I can see how that compares. Currently having fun getting 
the tablet to work, recompiling Qt at the moment...

Adrian