[Qt-interest] Xorg performance

[Clemens Eisserer]

Hi,

> I think you have no idea on how to measure the real life performance of
> an application.

Please not again that fanboy stuff.
This user has problems with QT on his (quite outdated) installation,
so I don't see a problem he is asking here.

> Your test, IMHO, shows nothing. Real life Qt
> applications do not have an "appaling" or "unusable" performance at all.

In real-life I see QT4 applications performing quite poor at resize
compared to QT3.
Most likely stupid backing pixmap allocation is the cause, would be
interesting to investigate further.

> I am not a specialist on this topic either, but perhaps you should
> re-read some of the old blogs of "Graphics Ninja" Zack Russin. He
> blogged about this kind of so-called benchmarks several times.

I guess the problem is QT4 is now double-buffered.
That means a lot of BLIT operations (drawing the backbuffer pixmap to
frontbuffer),
which is quite slow when its not hardware accelerated (like its the
case for XAA).
Is turning of double-buffering an option for you?

Sad you are not on Linux. For Linux there is sysprof which exectly
tells you where how many cycles are burnt.
Maybe there are some compareable "whole-system" profilers for your system?

Good luck!

- Clemens