[kde-edu]: KTurtle 4 performance

[Mauricio Piacentini]

Jeff Shipley wrote:
> Most stuff seems to run fairly quickly, I can draw a huge square spiral
> covering the entire canvas in a few seconds.
> 
> Speeding everything up (at least for the full-speed execution mode) is
> going to be a priority for me for the 4.1 release.

I think you will find out that a large part of the slowdown (specially 
if you have thousands of lines drawn) might be caused by our current 
approach to graphics. Each line is a new QGV item. This can give us 
several benefits, as we can scale the canvas to fit the size of the 
window like it is now, and antialias will be maintained. The effect is 
that now the "units" of measurement for the canvas are virtual, they do 
not correspond 1:1 to pixels on the image or screen, and this is very 
nice imo. It also enables other features like SVG export, maybe?
But when you have thousands (or maybe millions) of lines, refreshing can 
be a bit slow. Do you think the slowdown is in drawing, or in 
calculating? I also did not exhausted the possible QGV optimizations 
related to management of dirty rects. These are supposed to improve even 
more with Qt 4.4, and the usage pattern you mentioned for the fractal 
sample seems to match exactly a corner case in QGV performance, outlined 
by Andreas at 
http://labs.trolltech.com/blogs/2008/01/08/accurate-update-regions-for-thin-qgraphicsitems/

Of course, it might be an issue with the parser and recursion as well, I 
am not sure. Feel free to research this, but let me know if you are 
ready to commit huge architectural changes, as I am planning to 
experiment with a different approach to graphics as well (drawing on a 
fixed canvas size pixmap instead of creating vectorial lines.)

Regards,
Mauricio Piacentini