Quote:
Just my guess: KK is saving one of three passes in the blitter c2p (the first one probably, as it replaces the bytes in a word) by writing the data in scrambled order; if you do it in a clever way, it does not take additional CPU time.
|
Exactly that. Long live the a7 register.
Quote:
As he called it a 'crazy programming trick' at some point, my bet would be on some additional use of the stack's word alignment, but I might be off completely. :-)
|
You're completely on point. All my rendering writes through (a7)+, prescrambling columns at the byte level. Of course the other byte is another column, 8px to the right.