There is alot of statistics in Keith's and my Usenix paper. See: http://keithp.com/~keithp/talks/usenix2003/ Note that the full traces can be examined; you can trivially zoom in on any part of the data you like, using the tools we developed, and generate statistics for the traces of various applications selected. The traces we have on line are for application startup: but the methodology can easily be extended to allow examining any part of application behavior of interest. InternAtom is the largest single offender (for GTK); but there is also alot of other low hanging fruit, for GTK, Qt, Mozilla, etc... Different toolkits are doing different stupid things; and in one case, we need a window manager hint to enable Moz to no longer have to do a round trip per widget. About 90% or more of the round trips are unneeded in the application startup case, in our opinion, having examined the traces. As the biggest offenders get resolved, then we'll go after the next level offenders, etc. It all looks quite straight-forward. - Jim -- Jim Gettys Cambridge Research Laboratory HP Labs, Hewlett-Packard Company Jim.Gettys@hp.com > Sender: forum-admin@XFree86.Org > From: Juliusz Chroboczek <jch@pps.jussieu.fr> > Date: 05 Jul 2003 22:37:54 +0200 > To: forum@XFree86.Org > Subject: Re: Snappy GUI response > ----- > KP> 1) InternAtom -- Qt interns a bunch of atoms at startup time in > KP> one round trip; Gtk+ doesn't bother. > > How many? > > Juliusz > _______________________________________________ > Forum mailing list > Forum@XFree86.Org > http://XFree86.Org/mailman/listinfo/forum _______________________________________________ Forum mailing list Forum@XFree86.Org http://XFree86.Org/mailman/listinfo/forum