[FFmpeg-devel] [PATCH] modification of the MMX H264 MC chroma functions to support RV40

[Mathieu Velten]

2008/12/23 Mathieu Velten <matmaul at gmail.com>:
> 2008/12/23 Michael Niedermayer <michaelni at gmx.at>:
>>> or use a bigger rv40_bias table and use a shift to find the >>3
>>> variant in the table ?
>>
>> yes
>>
>
> I still don't know how to do.
> I don't want to use the rv40 table inside dsputil_h264_template_mmx.c
> if possible to do not pollute h264 code with a rv40 table.
>
> I can use a bigger table before the call of the function and I pass
> rng_reg and rng_reg_div_8 as arguments, but I am not sure it will be
> faster.

I don't think a single instruction which is only for the 1 dimensional
case and which is only execute once per function call (not in the for)
is a big deal, and it will be a mess if I try to use a table.

attached a cleaner patch (remove non related code, I will add this
code in a much proper way in a later patch).

I ported the h264 qpel mmx optimizations to the rv40 decoder, I will
submit a patch as soon as this one is accepted.
about 23% overall speed-up for my trunk against the current svn :)

Mathieu Velten
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rv40_mc_mmx_v7.diff
Type: text/x-diff
Size: 11982 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20081225/99330d60/attachment.diff>