Right now a huge number of offscreen buffers are created basically your grabing and painting the whole screen each time you draw. Its a bit better then that but not much. The biggest evil is the cairo font handling code. Considering the number of known major performance problems I think the results are encourging.
I'd suggest if your intrested in perforamance first fix the font impl and cairo and please send me the patch next if your running 16 bit color cairo needs support for 16 bit surfaces turned on. Finally if you look I'm using a ARGB surfaces as the default surface. And this is stuff I know is bad I've not tried to actually run any performance tests and don't plan to till I remove the worst cruft. 5 times slower is about right for the known problems so if there fixed it should be close to the X code with a bit more work to take advantage of any accleration provided by directfb it should beat it handily. I'd suggest if you want to understand the performance test DirectFB vs X11 Cairo/DirectfB vs Cairo/X11 and finally Gdk-directfb vs Gdk/X11 That should highlight were the major performance problems are looking at Gtk vs Gtk is not meaningful now. I think you will find its in the Cairo/Directfb implementation for the most part with some more problems in the Gtk layer. Also not in your test your also seeing optimization problems in Cairo even on X if this was done on arm with no fpu you probably hit some know performance problems and if your intrested in Cairo performance you need to follow cairo cvs closely and not for now use a 16 bit destiniation. Mike On 5/2/06, 강인표 <[EMAIL PROTECTED]> wrote: > > > > I tested GTK+ over DirectFB's performance using gtkperf tool. > > Test result surprised me. > > Below is my test result using $./gtkperf –a –c 100 in my target. > > > > GTK+2.6 over Tiny X : 97.14 sec > > GTK+2.8 over Tiny X : 160.28 sec > > GTK+2.9 over DirectFB : 458.65 sec > > > > I know GTK+2.8 is slower than GTK+2.6. > > But I cannot understand why GTK+ over DirectFB is much slower than GtK+ over > TinyX. > > Many reports said that DirectFB is slower than TinyX, > > but they didn't explain the reason of bad performance. > > > > If anyone know the reason, give me some information please. > > > > Best regards… > > > > ================================= > > > In-Pyo Kang (姜寅豹) > > > S/W Lab.4, Mobile R&D Group 10 > > > Samsung Electronics > > > --------------------------------- > > > Tel : 82-31-279-7675 > > > Fax : 82-31-279-6777 > > > Cel : 82-11-9825-7504 > > > E-mail : [EMAIL PROTECTED] > > > ================================= > > > > > > _______________________________________________ > directfb-dev mailing list > [email protected] > http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev > > > _______________________________________________ directfb-dev mailing list [email protected] http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev
