Rupert Pigott <email@example.com> wrote:
| Pascal Bourguignon wrote:
| > In anycase, at the time the Macintosh appeared, there were
| > already 680x0 based unix workstations.
| It was specifically the 68000. Fixes were made that took effect in the
| 68010 and 68020. Dunno about 68008. IIRC the problem was that you could
| not restart some instructions properly. Some UNIX workstations did use
| 68Ks, there was an Apollo that had two of them running in lock-step,
| with one of them one instruction behind the other. When the leading CPU
| barfed, action would be taken and the other CPU would take over. Someone
| in comp.arch worked on the Fortune boxes and IIRC he claimed they had a
| more elegant single CPU solution.
That would have been me, in <news:R4OdnfWrcMv1ekHdRVn_iw@speakeasy.net>,
replying to John Mashey. We tweaked the C compiler's calling conventions
enough to allow automatic stack growth by faulting off the end of the
stack to work reliably. See the referenced article for more detail.
But as I finished there:
Though there were certainly other places where the mc68000's imprecise
exceptions left no choice but to blow the offending process away...
Rob Warnock <firstname.lastname@example.org>
627 26th Avenue <URL:http://rpw3.org/>
San Mateo, CA 94403 (650)572-2607