* using MOVQ_WONE/MOVQ_BFE as two instruction instead of static memory value access
as its always faster
* PAVGB_MMX macro is using now mm6 -> mm7 is unmodified
* replaced original pixels_xy2_mmx with new faster and equal implementation
* replaced usage of mm7 for other then ZERO contstant in _rnd & _avg file with mm6
Originally committed as revision 632 to svn://svn.ffmpeg.org/ffmpeg/trunk
* calculating average in parallel to use two MMX units
* disabled xy2 put code - it's producing rouding errors
Originally committed as revision 626 to svn://svn.ffmpeg.org/ffmpeg/trunk
and being compiled in the same way as _avg.h
* PAVG_MMX macros accept also output parameter
* implemented faster put_pixels_xy2, but it has slightly smaller precission.
But there is not visible difference in the image quality - might be eventualy
easily switched back (#if 0 #endif)- please check
Originally committed as revision 624 to svn://svn.ffmpeg.org/ffmpeg/trunk