[theora-dev] Re: optimization for OLPC

[Chih-Chung Chang]

On 3/14/07, Ralph Giles <giles at xiph.org> wrote:
> Chih-Chung, I assume you're using libtheora 1.0a7. Could you re-run the
> decoding tests with dump_video from http://svn.xiph.org/trunk/theora-exp/ ?
> That should be instructive.
>
> Can the camera hardware generate YUV?
>
>  -r
>

Hi,

Here is the result for theora-exp:

sample               size         XO      Athlon
-------------------- ---------  -------- --------
320x240              320x240    56 fps   444 fps
322x242_not-divisibl 336x256    48 fps   387 fps
Building_On_The_Past 320x240    61 fps   571 fps
Elephants_Dream_1024 1024x576    4 fps    48 fps
Elephants_Dream_512- 512x288    18 fps   206 fps
MSD_ORACLE.vlog_2    368x240    85 fps   666 fps
chained_streams      320x240    56 fps   444 fps
monday_0930_Welcome  320x240    75 fps   599 fps
multi2               320x240    31 fps   249 fps
pixel_aspect_ratio   352x288    23 fps   272 fps
romalcc              352x288    73 fps   571 fps
timeskew1-t2s1       320x240    51 fps   521 fps
videotestsrc-720x576 720x576    12 fps    89 fps

It seems to be faster on only two cases, the 1024x576
clip and the pixel_aspect_ratio clip on Athlon.

Regards,
Chih-Chung Chang