[cairo] Text Grid Performance

Gerdus van Zyl gerdusvanzyl at gmail.com
Tue Feb 5 10:42:36 PST 2008


Ok, thanks for the feedback. Firstly I am on windows (windowsXP),
2.4ghz amd, ati radeon 9500. I am using the win32 backend.

In the text-scroll.py program Owen attached I get: 100 frames in
7480.99994659 ms, 13.3671969942 FPS, which is fairly consistent with
my previous results considering event handling/reflow overhead.

The clip is a rectangle and so far I know device-pixel aligned (I
don't use scale and all x,y,width,height is int). Tested
identity_matrix but made no difference.

Top functions by cumulative time is:
ncalls  tottime  percall  cumtime  percall filename:lineno(function)
67    1.280    0.019   20.128    0.300 platformwin32c.py:384(OnPaint)
67    0.399    0.006   18.848    0.281 uxpwindowapp.py:1219(event_paint)
6466/66    2.674    0.000   16.366    0.248 uxpwindowapp.py:1038(_renderone)
6334    6.415    0.001   10.892    0.002 uxbase.py:1079(showText)
6334    0.717    0.000   13.226    0.002 uxlabel.py:15(renderctx)
The rendering flow in my program is OnPaint -> event_paint ->
_renderone(x250) -> renderctx -> showText. Which to my eye looks like
cairo being the bottleneck.

Thanks.
~Gerdus


On Feb 5, 2008 7:49 PM, Vladimir Vukicevic <vladimir at pobox.com> wrote:
>
> On Feb 5, 2008, at 7:10 AM, Gerdus van Zyl wrote:
> >
> > My text rendering for each element is basically: save,clip,render text
> > using pango,restore
>
> Just off the top of my head -- is the clip a rectangle, and if so, is
> it device-pixel aligned?  If not, that could account for the
> performance that you're seeing.
>
>      - Vlad
>
>


More information about the cairo mailing list