theora dev - Oct 2007 - Blog post about Theora and MSVC assembly

Hello!


I wrote a blog post requesting help for Theora
on MSVC-syntax assembly. This is my first time
speaking for Xiph instead of just myself so
please teach me what to do better next time.
(Better offlist probably.)

My post might contain a few interesting bits for
Xiph members as well. It can be found here:

http://blog.hartwork.org/?p=86



Sebastian

I got it in my feed earlier and was impressed. That's awesome!
Have you talked to illi at all?

Silvia.

On 10/3/07, Sebastian Pipping <webmaster@hartwork.org>
wrote:
> Hello!
>
>
> I wrote a blog post requesting help for Theora
> on MSVC-syntax assembly. This is my first time
> speaking for Xiph instead of just myself so
> please teach me what to do better next time.
> (Better offlist probably.)
>
> My post might contain a few interesting bits for
> Xiph members as well. It can be found here:
>
> http://blog.hartwork.org/?p=86
>
>
>
> Sebastian
> _______________________________________________
> theora-dev mailing list
> theora-dev@xiph.org
> http://lists.xiph.org/mailman/listinfo/theora-dev
>

Sebastian Pipping schrieb:
> Hello!
>
>
> I wrote a blog post requesting help for Theora
> on MSVC-syntax assembly. This is my first time
> speaking for Xiph instead of just myself so
> please teach me what to do better next time.
> (Better offlist probably.)
>   
Sorry to answer to such an old thread, but anyway...

I just had a view at the new trunk and it seems that noone has 
volunteered to do the porting job yet. I'm a old assembly low-level guy 
and I do most of my work on the windows OS, so I'm able to do the job in 
theory.

There is not that much MMX and SSE stuff in the trunk so it should be a 
quick thing to do. Maybe 2 to 3 days.

However, I don't like the idea to write that much specialcase code just 
for a single compiler. This is even worse than branching since you can't 
simply merge two different compiler extensions back to a unified codebase.

I'd like to convert the code into NASM-format. That way linux, windows 
and all other x86 platforms I can think of are able to compile from the 
same codebase and get the benefit of the SIMD implementations.

Drawback is of course yet another dependency for the build-process. 
However, NASM is easy to install. Just put the binary somewhere in your 
path and you're done. No configuration or so. Last time I've used it it 
was a stand alone executable.

Any feedback welcome,
  Nils