18

KDE 3.5.10 is extremely snappy and responsive for pretty much everyone however KDE 4.1-4.2 is quite snappy and responsive for some and not so for others. Obviously those people for which the performance is good, have the same KDE code base as others however their versions of QT and / or Xorg / graphics card drivers may vary.

It would be nice to have some sort of test which users can run which has a look at the actual drawing operations to see which operations are slow on a given configuration and if possible where to report bugs and / or look for solutions.

For example I have 2 machines, similar spec, both with Intel hardware, running different versions of Xorg. One runs KDE 4.2.1 at a reasonable usable speed, the other does not. Now I know the fault lies somewhere outside KDE however I can't submit a bug report as i have no idea where the problem is likely to be and i have no metrics to present to developers. I can't write my own tests as I have not the graphics experience in order to know if a particular function actually draws something on the screen or just writes to a buffer somewhere.

Both KDE3 and XFCE run fine on both machines so I think it's down to certain function calls and situations. It would be nice to be able to identify these so that people can at least track them.

Excellent idea. This could be used to correct inefficient paths of code by the developers of the drivers concerened.

Yes, excellent. Glxgears isn't doing what people use it for. But the fact that people keep "testing" with it shows that tests are necessary. Having KDE-specific tests (e.g. drawing different windecos) would help a lot to find problems (and then maybe be able to convince the xorg and driver devs to do something about it). Great idea!