13 May
2011
13 May
'11
4:44 p.m.
Sheesh!
Any OS needs device driver support the serial IO is just plain damn trivial.
Mass storage is the PITA. However if your building the machine that
problem
be comes one of implementation and since the implementer is the likely
programmer where the problem? Use something dirt simple like IDE or CF
that does 16 bits and is buffered so PIO is the easy out.
As to OS, Anything fits if you use a good compiler and don't try to force
PCisms into it. Also you have working platforms like real Qbus or
Ubus 11s to develop on.
On 05/13/2011 11:21 AM, Dave McGuire wrote:
Andrew, you're awesome, but you really, really, REALLY need to switch
to a real mail program. Your reply format is...well, painful at best.
No offense is intended. Please see below.
On 5/13/11 10:52 AM, Andrew Lynch wrote:
Sheesh!
Any OS needs device driver support the serial IO is just plain damn trivial.
Mass storage is the PITA. However if your building the machine that
problem
be comes one of implementation and since the implementer is the likely
programmer where the problem? Use something dirt simple like IDE or CF
that does 16 bits and is buffered so PIO is the easy out.
As to OS, Anything fits if you use a good compiler and don't try to force
PCisms into it. Also you have working platforms like real Qbus or
Ubus 11s to develop on.
Of course if you never programmed an 11 then you completely lost as PCisms
are not there. You have a clean CPU no mmu for T11 but its possible to
copy
the 11/23-11/34 MMU with not so much work. And you have a full spec and
docs available for any 11. What doesnt fit in a 8086 did fit well in
PDP11 due
to it being truly CISC without bogus 8080 registers.
Netbsd, even uClinux are out there and would fit. Just learn how to do
real
virtual memory and swapping. The KT11 MMU made that programatically
possible.
Allison
Also, I was very surprised to learn there are no
free/open source
operating systems for the PDP-11 which would be required to adapt for
existing S-100 peripherals.
Apparently not even NetBSD is available which is stunning as it runs
on practically every CPU in some form AFAIK. The lack of free/open
source operating systems is a major problem since it then requires
all the PDP-11 peripherals to be hardware compatible. As a result,
either the entire PDP-11 CPU with all peripherals fit on a single
S-100 board or a suite of dedicated boards would be needed. Neither
scenario is realistic IMO.
Well, there's always 2BSD. It actually runs quite well, is pretty
solid, and supports reasonably modern networking. It's not a bad OS.
NetBSD, while wildly portable, would be damn near impossible to port
to the PDP-11. It's just too "big". Don't forget that it's a
16-bit
architecture with (at most) 22 bits of address space. Even 2.11BSD,
the most modern iteration of the 2BSD lineage, took quite a bit of
shoe-horning to finish.
Hardware compatibility of emulated or re-implemented PDP-11
peripherals shouldn't be that much of a big deal. Keep in mind most
of these peripherals were designed 30-40 years ago. The designers of
the time were working with far tighter constraints than we have now,
and to top it off, were almost always working on paper. I've studied
most of the peripheral controller designs and the majority of them are
very simple. Pleasantly so, I might add. :-)
For those who will point to MSCP controllers and disagree with that
assertion, I will point out that most DEC-built MSCP controllers
(RQDXn family) used a T-11 processor, not exactly a high-powered
machine. Many third-party implementations were built around
mcs51-architecture microcontrollers. This places fairly low limits on
complexity.
-Dave