> huge improvement; every rendering pass after the first one would be> twice as fast!
I think this should only be 33% as fast if I understand it. For each new
pass you would calculate 3 new pixels + reuse one old one for every square
of 4 pixels. 4/3=1.33
-tgq