Re: gtkperf regression and gtknotebook

["Padraig O'Briain" <Padraig Obriain Sun COM>]

Padraig O'Briain wrote:
> Matthias Clasen wrote:
>   
> > On 7/5/06, Padraig O'Briain <Padraig Obriain sun com> wrote:
> >     
> > > I am not sure that cairo has anything to do with it; XRenderComposite is
> > > being called from draw_with_pixmaps in gdk/x11/gdkdrawable-x11.c.
> > >
> > > Since I sent the original mail, I have noticed,when using gtk 2.8, a
> > > call to gtk_paint_flat_box
> > > with what looks like an abnormally large value for width. I need to
> > > follup up that lead.
> > >       
> > Any news on this, Padraig ?
> >     
> Not yet.
>
> I have noted some differences between gtk 2.4 and gtk 2.8 but I have no 
> evidence that they have anything to do with the time difference between 
> gtk 2.4 and gtk 2.8.
>
> The differences I have seen are:
>
> 1) Size allocations for some widgets are slightly different; i.e. differ 
> by 1 or 2 and seems to be pango related.
>
> 2) The widgets in GtkComboBoxEntry have changed ; see gtkcombobox 
> version 1.168 and gtkcomboboxentry version 1.16.
>
> 3) Implementation of gdk_window_invalidate_maybe_recurse has changed in 
> gdkwindow version 1.170.
>
> 3) GtkStatusBar changed in version 1.46 of gtkstatusbar.c.
>
> Padraig
>
> _______________________________________________
> Performance-list mailing list
> Performance-list gnome org
> http://mail.gnome.org/mailman/listinfo/performance-list
>   
As I seemed to be getting nowhere I started changing gtkperf to see if I 
could figure out that way why the performance using gtk 2.4 and gtk 2.8 
was different.

The program gtkperf UI consists of a GtkNotebook with 6 pages. The first 
page is used to control the tests when run manually and the other five 
pages contain the tests.

The test I focused on was GtkComboBoxEntry as that seemed to have the 
worst degradation. This test is in second page. I removed the 
GtkComboBox, GtkProgressBar and GtkSpinButton widgets from that page. 
This did not have any noticeable impact on that time taken to run the tests.

On gtk 2.4, I tried removing the other pages from the GtkNoteBook. 
Removing the third to the sixth pages and their tabs did not make a 
noticeable difference to the time taken.

Removing the widgets in the first page improved performance by about 1 
second, from 8.7 to 7.7 seconds.  The improvement all seems to be 
associated with removing the label called "Start".

I noticed that if I then replaced the gtk-dialog-info image in the first 
page's tab with a text label that the time taken to run the test was 
reduced by about 5 seconds to 2.7 seconds.

If I then add back in the widgets on the first page up to and including 
the label called "Start" the time taken goes back up to 8.7 seconds.

This has left me very confused as it looks like adding or subtracting 
widgets from a page which is not seen on the screen impacts the performance.

Any suggestions as to what is going on or how to figure out what is 
going on welcome.

Padraig