Hi, I've upgraded your patch against GTK+ 2.10.6, and we are getting great performance figures compared to GTK+ 2.10.6 with pangocairo too (basically at the level of GTK+ 2.6.10 again). Right now I'm working on a python/cairo script to get some nice graphics from a torture test session with several GTK+s, hope to get it ready soon. Xan El jue, 19-10-2006 a las 15:04 -0500, ext Matt Hoosier escribió: > Hi, > > Attached is a patch against gtk+-2.8.20 CVS which replaces the > pangocairo text rendering with the old pangoxft code from 2.6. It > targets embedded systems which currently suffer some performance > bottlenecks using Cairo for text. > > Note that in a few cases some fields of structures (e.g., > _GdkGCX11.fg_picture and _GdkGCX11.fg_picture_color) had to be > re-added. This was done mainly because the restored pangoxft code used > them to calculate things. I'd gladly accept criticism about betters > ways to fetch the information using normal 2.8 internal state rather > than re-add those fields. > > gdk_screen_get_font_options() and gdk_screen_set_font_options() also > don't do anything now. I'm not so sure how much a problem this is in > practice. > > Using Manu Cornet's gtk-theme-torturer on some ARM hardware, 2.8.20 > originally took about 55% longer than 2.6.10 to complete the GtkLabel > run. After these patches, it takes only about 8-10% longer than > 2.6.10. > > Thoughts? > _______________________________________________ > Performance-list mailing list > Performance-list gnome org > http://mail.gnome.org/mailman/listinfo/performance-list
Attachment:
pangoxft2.10.6.diff.gz
Description: GNU Zip compressed data