Re[2]: [XviD-devel] Comparison "old" vs "new" adaptive quant code

[Radoslaw Czyz]

Hi

> Even better would be if you did some tests with ADVANCED_DIAMOND.
> If it doesn't show any problems (why should it? I don't know), we could
> even make it default for all quality levels. In theory, it should be
> better quality _and_ faster than "usual" diamond.

Well.. I don't expect it to be faster. Nope. In theory I could imagine
a search pattern when advanced diamond would be faster (it doesn't do
any double checks) but... well OK, test it to be sure.

@gruel: do you remember the time when I wrote advdiamond? Before I did
it, I tried another 'diamond'. That diamond was much faster but
conducted less checks - it lowered PSNR a bit. In the ME code I sent
you, 'normal' diamond is exactly this code. It's fast, advdiamond is
better, square is better again.

Radek