Re: MMX speedup for Floyd Steinberg error diffusion

On May 8, 11:00 am, Terje Mathisen  <spamt...@[EMAIL PROTECTED]
> wrote:
> How does it handle the corner/edge cases?
>
> I.e. the last pixel on a scan line has no down-right neighbor, the first
> pixel has no down-left and the last scanline has nothing below at all!

The "off-screen" pixels as ignored, because they do not contribute to
the error information distributed to the visible pixels.  A dumb but
easy way to implement this without conditional branches every
iteration (to check if you are operating on off-screen pixels) is to
operate on a bitmap that has a 1-pixel border around the right/left/
bottom sides.

> Anyway, to speed up the dithering stage I would use MMX/SSE and work on
> multiple lines in parallel, keeping the intermediate results in
> registers until we're finished with them.

I don't see how you'd implement that, because the very top-left pixel
can conceivably promote enough error, distributed downward, to affect
the bottom-right pixel...