>I havn't said that GObject signal marshalling is a major bottleneck. >I've said about overal run-time design of gtk... you said "run time binding to signals". >I have the following manual mesurments, wich are very inaccurate... >It's near 3 seconds betwean clicking on gnome-terminal launcher and the >appearance of the window... >It's near 5 second to start mozilla-thunderbird... >Easy-tag (gtk1.2 application) starts very quickly --- not noticable... so there is some evidence that GNOME might be slow. we also know already that mozilla apps are large and heavy, and that they don't even use GTK directly. this has very little to do with GTK. >What can I do to improve it? no idea, but the examples you gave are a long way from suggesting that there is something deeply wrong with the basic "runtime" design of GTK. the gnome-terminal is not one of the faster terminal emulators anyway, and thats not because it uses GTK, but because of other aspects of its design. registering apps on startup with "The GNOME Desktop" is definitely slow, and can lead to lagginess in application startup. But for a fair comparison, try the GIMP - a big badass program written in 100% GTK: it shows almost immediately *and then* appears slow as it churns through its non-GTK-related initialization process. --p _______________________________________________ gtk-list@xxxxxxxxx http://mail.gnome.org/mailman/listinfo/gtk-list
