kiessig@sri-unix (08/02/82)
Regarding the previous article on differences between the 68000 and the 68010: You should already know that the 68000 does not handle instruction restart. The only way you can implement a true Virtual Memory architecture using the 68000 is to use two processors. One does the real work, and the other handles faults (if you have a page fault, the second processor reads in the page for you, being very careful not to fault itself, of course). The second processor does not have to be a 68000, but it's cleaner if it is (Apollo has taken this approach, as have a few others). The disadvantage here, of course, is that the first processor can do nothing while the second one is handling the exception. You can optimize this a little, but the code is very nasty (speaking from experience here). The 68010, then, needs no fancy hackery. Like a PDP-11, it can restart instructions. Equipped with an MMU, you could handle various forms of VM. The SUN machine has an architecture which lends itself nicely to dropping in a 68010 (pin-for-pin compatible with the 68000, remember), and a little exception handling code, and ending up with a much more powerful system. Seems like this has been explained before, somewhere. I can only hope the readers of this list are now enlightened. Rick Kiessig