Re: [Gnash-dev] Fwd: Teddymark benchmark for Gnash

[strk]

On Thu, Jan 08, 2009 at 11:15:02PM +0800, Carlos Nazareno wrote:

> It's been uploaded and the urls below should run now.
> 
> >> http://www.phlashers.com/projects/teddymark/teddyMarkAS2_100.html
> >> http://www.phlashers.com/projects/teddymark/teddyMarkAS2_150.html
> >> http://www.phlashers.com/projects/teddymark/teddyMarkAS2_200.html
> >> Which have 100, 150, and 200 Teddies running around the 600x400 stage.

I get 3.0 to 4.1 FPS reported with Gnash using AGG renderer
on an AMD Athlon(TM) XP 2000+ with 3341.17 bogomips.
CPU bound.

It's interesting to note that the reported
FPS keeps slightly growing, but the same
happens with adobe player 10 (which starts
from 10.0 FPS and goes up).

With a profiling build of Gnash I've verified
the most time is spent in rendering shapes: ~38.5%
followed by 3.65% for malloc and 2.66% for PropertyList::getProperty.

In particular these are the first two entries:

 Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total
 time   seconds   seconds    calls  ms/call  ms/call  

 26.91      7.38     7.38  2513521     0.00     0.00 
 gnash::agg_style_bitmap<...>::generate_span(agg::rgba8*, int, int, unsigned)

 10.61     10.29     2.91    75921     0.04     0.15 
 agg::render_scanlines_compound_layered<...>(...)

>From the call graph it seems the first one (generate_span) is called
by the second one which is called from 
void gnash::render_handler_agg<>::draw_shape_impl<..>().

--strk;