[Ffmpeg-devel] swscale

[Marc Hoffman]

I guess I was looking into swscale again and sorry guys but this code
is a bit frustrating. The infastructure that the codes sit in are just
so hacked up at this point its pretty hard to add new architecture
specific stuff.  I think originally its intent was to keep things nice
and clean and have each architecture have its own set of scaling,
color space converters and resamplers.  But at this point its a bit
indecipherable with all the complex macro expands and renaming stuff.
Not that this matters the code is so complex that it locked my EMACS
session up.

Does anyone have some direction here on what would make the most sense
for moving this forward?  I guess I could just follow along and ifdef
stuff up but in my opinion that would just make a bigger mess when I
add the Blackfin backend.  There is so much good engineering in here
and I guess following along with the same adds as the altivec stuff is
the path I'm on but if there are any other ideas I wouldn't mind
listening.

Marc