[Kde-games-devel] QGV repaint and update performance

[Mauricio Piacentini]

Mauricio Piacentini wrote:
>> However, why this is is totally unclear to me. This must be some QGV problems.
>> Maybe just hope for a new Qt release. Does anybody know whether Trolltech is 
>> aware of these issues? 
> 
> Yes, they are. See my reply to Tomas, we discussed and tried to address 
> this before last aKademy, and Andreas (from TT) even looked at 
> KGoldRunner specifically in Dublin.
> already addressed. Maybe you could try with a 4.3 snapshot, although I 
> am not sure which version will be used by KDE4.

Just a disclaimer: I am of course not 100% sure at all that the current 
issue we are seeing in this case (KWin4) is the same we observed in 
KGoldRunner back in Sept/Oct, just that it seems to be. It may very well 
be that this is already solved in Qt4.3, or that we happen to share a 
bug in the game implementation that was causing that problem and this 
one as well. I unfortunately do not have time to help you investigate 
this right now, as I had my share of QGV hacking/debugging last year 
already, and I am currently trying to finish the KMines redesign. What I 
can tell you is that Andreas tried to address the KGoldRunner slowdown, 
but we could not solve it at that time with his patches. The process is 
documented in the qt4-preview archives, with test files and benchmarks.

Also please notice that the KBounce issue is a different one, and has to 
do with the lack of partial invalidation of the background. It is more 
of a design issue that causes the performance hits. We also found this 
problem it in one of the several QGV-enabled iterations of KGoldRunner.

In other words, I am just writing about the profiling we did a few 
months ago, and our attempts to solve these problems, in the hope it 
will save you guys some time. It might be a QGV limitation, it might be 
a mistake in our porting code, or it might be something different.

Regards,
Mauricio Piacentini