Re: gtkperf results

["Clemens Eisserer" <linuxhippy gmail com>]

Hi again,

> Attached is the output of gtkperf in 4 configurations:
>
> Radeon 9800 w/ FireGL drivers - GTK 2.9.4 - Unpatched
> Radeon 9800 w/ FireGL drivers - GTK 2.9.4 - your patches
> Intel i915 - GTK 2.9.4 - Unpatched
> Intel i915 - GTK 2.9.4 - your patches

Well both the ATI card powered by the FireGL drivers and the
Intel-Onboard chip show no improvemnts when caching the pixmap -
therefor its really not worth the roubles so I wrote NVidia that this
is a problem and they should investigate ;-)

However I've still doubts wether its a good idea to create that many
pixmaps - maybe the code could be modularized and adopted for
different needs. Creating only one pixmap (largest expose-region) and
share it across sub-windows would be interresting - however I guess
its not that easy keeping the a bit hackish style the double-buffering
mechanismn is realized (maybe I am still too much in the OO world).

I'll continue to write some gtkperf scenarios which will simulate
situations where I saw problems - maybe really bad numbers could
motivate guys here to investigate a bit more - since performance stuff
still seems to get ignored more or less (hey I got an email from Owen
- unbelievable *lol*).
I'll also try to write QTPerf which will do exactly the same in QT
(ok, comparing toolkits is  stupid, but who descides whats slow and
whats fast) to see how good/weak we are compared to other major
toolkits.

Thanks for all the patience and help :-)

lg Clemens