[XviD-devel] 15% faster search16, if anyone is interested

[Michael Militzer]

Hi,

On Fri, 12 Jul 2002, Radoslaw 'sysKin' Czyz wrote:

> Hi
> As some of you know, I'm trying to create a different inter4v motion
> search.
> As a by-product of my experiments, I've discovered that four sad8s
> conducted in place of one sad16 lead to about 15% faster encoding (for
> non-inter4v). Of course this is not because 4 x sad8 is faster, but
> because it's possible to do an early termination after each sad8.
>
> I was a bit surprised about this, but whatever, It doesn't matter much
> for my code - it will only benefit if inter4v is off.
>
> The results were made with pmvfast, no-inter4v, advdiamond, halfpel
> refine and ext_search. It used xmm sad-code (for AthlonXP.. it's xmm
> right? It doesn't matter either)

hm, this is a big surprise for me. We had some early termination code
within the sad16_xmm code itself earlier, but we removed it, because the
code became slower instead of faster.

At least on my Athlon 900 the xmm code was definitively slower with early
termination than without and it should be even worse for faster computers
(your XP)...

could someone else verify sysKin's results please?


bye
Michael