Quote:
Originally Posted by Samurai_Crow
Since every longword in the AnimOb structure except a minimum list node at the beginning is misaligned causing the speed penalty to be compounded to such a wild extent that I'd say such a design only works properly on 16-bit Amigas. (Although adding a padword to the beginning of the structure before you allocate it would be a quick and easy fix, I've been told that the Animate() subroutine is only about 250 bytes of code and easily duplicated.)
|
I see. You speak of longword data accesses that are only word aligned. The 020+ handles it better than most processors. It doesn't hurt performance that much unless the data is accessed inside a loop. I believe the 040, 060 and N68k processors will have no penalty for these types of accesses in the cache either. It's less than optimum but I've seen worse performance killers.