it has some problems rendering single slices. It looks like it is
off by one scanline somewhere.

Some notes on things that need to be done with yuv2rgb_mmx.c

- fix the major bug first
- merge the different combinations of depth/colour order into one
  function body + macro/inlines. See yuv2rgb by walken for a good
	example of this. I'm not maintaining 6 different versions of a fifty
	line mmx asm routine :)
- change the assembly to use the mmx.h style assembly.

Thanks to Ollie too, for being so patient.

cheers,
aaron