[XviD-devel] Qpel Problem Samples

[iibot at gmx.at]

> > I know this was meant as a joke, but I believe things won't slow down
> much
> > if the first and second pass would be interleaved, since they work on
> the same
> > dataset per frame/slice/whatever anyway.
> 
> Hm, I'm not so sure about this, because only the incompressed input
> picture could be used by both tasks, everything else would almost
> immediately differ. No motion vectors etc. could be reused. for the same
> reasons as with "hinted MV". 

I guess you're right on this one.

> But on a P4 with Hyperthreading or a SMP system, encoding the same clip at
> the same time with and without QPel would of course be faster than one
> after the other. 

That reminds me of a question I wanted to ask for a long time:
Do you use software prefetch for motion estimation? I know that P4 (and AXP)
HW prefetch is inappropriate in this case.
Or did you save that for final optimizations on 1.0?
Maybe it doesn't work at all, due to limited amount of buffers (outstanding
memory load)?

bis besser,
Tobias

-- 
COMPUTERBILD 15/03: Premium-e-mail-Dienste im Test
--------------------------------------------------
1. GMX TopMail - Platz 1 und Testsieger!
2. GMX ProMail - Platz 2 und Preis-Qualitätssieger!
3. Arcor - 4. web.de - 5. T-Online - 6. freenet.de - 7. daybyday - 8. e-Post