theora - Sep 2009 - Taking advantage of SIMD

After creating a simple decoder for theora using liboggz i started profiling. As
someone else stated in the lists the conversion from yuv 2 rgb takes a
significant time. I decided to go on the way of SIMD not shaders because SIMD is
supported on a wider range of computers. I managed to strip out the yuv 2 rgba
conversion code from liboggplay but it seems i am missing something. It runs but
the image is distorted, like this:

http://www.gamesrebirth.com/adrian/graphics/distorted.JPG

The code looks like this(only a few lines):

http://pastebin.com/m185d7b76

Thanks,
?? Adrian



      
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.xiph.org/pipermail/theora/attachments/20090904/d5121aee/attachment.htm

On Fri, Sep 4, 2009 at 6:59 PM, Manolache Adrian<prog_ady at yahoo.com>
wrote:
> After creating a simple decoder for theora using liboggz i started
> profiling. As someone else stated in the lists the conversion from yuv 2
rgb
> takes a significant time. I decided to go on the way of SIMD not shaders
> because SIMD is supported on a wider range of computers. I managed to strip
> out the yuv 2 rgba conversion code from liboggplay but it seems i am
missing
> something. It runs but the image is distorted, like this:
>
> http://www.gamesrebirth.com/adrian/graphics/distorted.JPG
At a guess, your row stride is off.  It will not be the same as the
image width * byte width.

Monty