Re: GtkPerf data for GTK+ 2.6 vs GTK+ 2.10 vs Maemo-GTK+ 2.6 (on a Nokia 770)

[Federico Mena Quintero <federico ximian com>]

On Wed, 2006-08-16 at 01:08 -0700, Carl Worth wrote:

> As I suggested above, the "Resize" column does look fairly stable
> across these results, usually differing by only a few percentage
> points, and differing at most by 10% in the 3 tests above.

I'm not sure how to get rid of the variations.  What happens if you
increase the number of iterations a lot, so that each test takes a few
minutes to run?

Also, I'm not convinced that GtkWidgetProfiler does the right thing for
all kinds of widgets.  It should be fairly accurate if you are testing a
GTK_NO_WINDOW widget, since is uses property changes on the toplevel
window to know when the server is done drawing, but it doesn't have any
special behavior for widgets that *do* have their own window.

That is, GtkFrame should be fine, but GtkEntry will be problematic.  I
don't know if GtkFrame *is* fine right now :)

Maybe GtkWidgetProfiler needs to dig down the window hierarchy in the
widget you pass it, and actually add instrumentation to each individual
window.  When it has received info that all of them have repainted, it
can print the final result.

Maybe S�has ideas about this...

  Federico