Hi,
> That and the broken text in the intro. it looks like it's not getting
> dithered/blended in properly. I think it was good before.
This is only visible if at least one of the scaling factors is even (it
doesn't look too great with two odd factors, though) and bilinear
filtering is enabled for text. I think we'll either need a different alpha
threshold or filtering which regards the diagonal neighbours as well
("trilinear"?). A combination of both would probably be best.
BTW, dithering during crossblitting (is that the correct term?) has been
removed; it had a noticeable performance impact at practically no visible
gain on my box (any observations on other systems?)
A "correct" crossblitter would be quite a bit slower than the current
implementation, except on accellerated hardware (which none of our current
drivers supports), or maybe with 8/8/8 bit layout (24bpp/32bpp) and
appropriately optimized functions doing parallel multiplications.
BTW^2, who made that CVS update? Please post your names with all commits
in
the future, as they're not implicitly put into the messages going to the
mailing list. TIA!
llap,
Christoph
