How will cairo influence GTK's performance?

[Clemens Eisserer <linuxhippy gmail com>]

Hello,

First of all please do not interpret this post as a troll post - it is
absolutly not my intention to blame anybody for the great work on gtk
or to spread FUD or whatever.
The following lines are just experiences of mine, so not valid in
general and not ment as troll stuff...

I am using GTK since a long time (both coding gtk applications and
using gtk based application) and I think GTK is a great toolkit -
however I am experiencing performance problems since GTK-2.0 and first
I thought this is just because tons of unoptimized code since so many
functions have been re-implemented.
However now we are at 2.6.? and I still do not see significant
performance improvements.

I also have seen several applications which are cross-toolkit or which
switched from gtk-1.2 to 2.x.y and they all perform better with
gtk-1.2 or the other toolkit.
The GTK2-port of SWT/Eclipse for example is the slowest port of all
swt ports (win32, mac, motif, fox, gtk2) and since mozilla switched to
gtk2 their interface is really sluggish on my Authlon-800 although I
have an 2d accerlated nvidia card.
The bug-report which (among others) mainly adresses GTK2 performance
problems is the report with the second most votes of all eclipse bugs:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=37683
They also profiled GTK and found out that for eclipse the problem
mainly lies in pango.
I already tried recompiling GTK not thread safe but it didn't help a lot :-(

I would be really interrested in helping to speed up GTK, however I am
a bit concerned wether tuning the x11 based backend does still make
sence now when GTK is switching to cairo as rendering backend. (are
there any numbers how cairo does influence gtk's performance?)
Are there any know performance problems where man-power is needed?

Thank you in advance, lg Clemens