21 Jun
2012
21 Jun
'12
4:50 a.m.
On 20 Jun 2012 at 19:41, Eric Smith wrote:
In the 16/16B/6000 systems, it was perhaps somewhat of
an advantage to
have a Z80 to serve as the I/O processor (although TRSDOS-16 failed to
overlap any processing), but it was also somewhat of a disadvantage in
that the 68000 could not access I/O devices directly at all, even in
cases where that might have improved performance. Another problem was
that the 68000 subsystem was designed to be able to interrupt the Z80
to request service, but because this only worked in machines sold as
the Model 16/16B/6000, and not in upgraded Model II/12 machines, the
software generally was not able to take advantage of it. On the other
hand, the Z80 could generate interrupts to the 68000 on all machines.
When we were working on our 80186-80286 box and were writing the I/O
routines for Xenix, I recall going up to Bellevue and being presented
with a listing (we used green-bar tractor-feed paper for everything
back then) for the I/O system for the Model 16 Xenix to use as a
model.
Given the I/O devices that the Model 16 was working with, the Z80 was
probably quite adequate to the task. Speaking from my own
experience, it really does make implementation of a protected-mode
operating system simpler when all you have to do is pass messages to
an I/O processor.
--Chuck