xf86-video=intel-1.99.92 performance problems and disabled DRM warning

[Pavel Troller]

Hi!
  Replying to myself:

> Hi!
>   I'm here with the first experiences with the RC2 intel driver.
> 
>   The most important is that it seems to have some performance problems in
> comparison with the old driver.
>   - During KDE session restoration, many windows are opened concurrently.
>     With the old driver, the startup melody was played smoothly. With RC2,
>     it's skipping badly.
>   - When switching virtual desktops, windows are redrawn more slowly than
>     with the old driver. I.e. a footprint of the window appears immediately,
>     but its contents is drawn a bit later. With the old driver, the windows
>     are drawn instantly, all at once.
>   - I have a superkaramba theme updating two graphs once per second. With the
>     old driver, this update has no influence to other screen operations; with
>     RC2, every app stops updating for a while once per second, until karamba
>     is killed. For example, glxgears goes from 1450 to 1250 fps due to this
>     and watching the video is almost impossible.
> 
> In the server log, I see warning about insufficient DRM version. Is it
> possible that this causes all the problems ? What component I have to update ?
> I have libdrm-2.3.0 and kernel 2.6.20. The log snippet follows:
> 
> (WW) intel(0): DRM version 1.6 older than required 1.7 for DRM memory manager.  Disabling.

Fixed this one. Kernel drm module from 2.6.20 is too old. I had to recompile it
from git drm sources. Now, the warning is over.

  However, there still seem to be some graphical performance problems.
Especially the superkaramba applet is still heavily degrading the graphical
performance, causing that once per second all, what moves on the screen, jerks
badly - starting from glxgears over xv video ending with card moving in
solitaire :-). I think it's not drm related. The applet is doing the
"artificial translucency", i.e. it's not using alpha, but reading the contents 
of the root window and manually compositing. Isn't it possible that this causes
problems to the new driver, while the old one was okay ? 

           With regards, Pavel Troller