[Gegl-developer] Hi, need any help?

[Jan Heller]

On 17:55, Mon 10 Mar 08, Martin Nordholts wrote:
> That looks very interesting. Do you think you could provide some
> benchmarking data of the performance improvements in variuos situations?
> It would be useful to have

Hi,

 I used babl/tests/babl_fish_path_dhtml as a simple
 benchmark. Here are running times of babl_fish_path_dhtml
 w/o and w/ the patch applied:

 jheller at voyager ~/projects/gegl/babl/babl/tests $ time ./babl_fish_path_dhtml > babl_fish_path_0.0.20.html

 real    0m10.463s
 user    0m9.441s
 sys     0m0.040s

 jheller at voyager ~/projects/gegl/babl-patch/babl/tests $ time ./babl_fish_path_dhtml > babl_fish_path_patch.html

 real    0m3.844s
 user    0m3.416s
 sys     0m0.048s

 Here are the resulting HTML files:

 http://www.ms.mff.cuni.cz/~hellj1am/WWW/babl_fish_path_0.0.20.html
 http://www.ms.mff.cuni.cz/~hellj1am/WWW/babl_fish_path_patch.html

 Such a dramatic performance gain is not to be expected from
 any practical usage of the babl library, since the database
 code is not as heavily used.

 A simple 512x512 16-bit/color RGBA PNG image load/save test
 using gegl gives following results for runs w/o and w/ the
 patch. I took the respective best results from several
 tries:
 
 W/O
 real    0m6.367s
 user    0m3.476s
 sys     0m0.168s

 W
 real    0m2.930s
 user    0m2.400s
 sys     0m0.096s

 I hope these results give a better idea about the performance
 of the changes.

 Regards, 
   Jan