[XviD-devel] asm-ed sad32v

[Radek Czyz]

Hi everyone,

I was wondering if one of asm programmers is bored? If it's the case,
would it be possible to write a fast sad32v function?

sad32v is currently only used in rrv-search, but of course that's not
the reason I'd like to see it. My plans are to change P/B/I decision
to evaluate 4 macroblocks at once, similary to inter4v search in
reduced resolution. This should be much faster, while lower precision
can be simply compensated by changing decision thresholds.

I hope it's not difficult to do ;)

Best Regards,
Radek

PS sad32v_c can be found in motion/sad.c in our development tree.
PS2: for my tests, xmm is OK, but sse2 isn't