Path: utzoo!utgpu!watserv1!watmath!att!pacbell!pacbell.com!mips!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!ames!haven!udel!mmdf
From: Pat_Barron@transarc.com
Newsgroups: comp.os.minix
Subject: Re: Once more. No more.
Message-ID: <20999@nigel.udel.EDU>
Date: 4 Jun 90 17:50:25 GMT
Sender: mmdf@udel.EDU
Lines: 63

Fred van Kempen writes:
>>From article <20666@nigel.udel.EDU>, by Pat_Barron@transarc.com:
>> "Vanilla" V7, straight from Murray Hill, did *not* run on the 11/34, since
>> the kernel was big enough that it required separate I&D space.  DEC had
>> a version they called "V7m", which had kernel overlays and could run on
>> non-separate I&D space machines like the 11/34.  And it had fsck, too.
>  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>Big mistake here: the '34 WAS A FULL UNIBUS MACHINE WITH 18-BIT
>ADDRESS LINES AND SEPARATE I AND D SPACE !!!!!!!

Sorry, this is not correct.  The 11/34 had (has) an 18 bit Unibus address
space, but no separate I&D.  The separate I&D machines are (from memory;
hopefully I won't leave out any):  the 11/44, 11/45, 11/53, 11/55, 11/70,
11/73, 11/83, 11/84, 11/93, and 11/94 (yes, there are such things as the
11/93 and 11/94 - Digital just recently announced these two new PDP-11's!).

>Thus, the 11/34 has (yes, present time.  I still have some!) an addressing
>capability of 256Kb (minus 8Kb UNIBUS space at the top).  One process could
>be split into two segments each: I and D.  This allowed for much larger
>processes to be run.

If you have separate I&D, you don't have an 11/34.

>Remember, guys, that all major PDP-11's (so, not counting the '05, '10
>(== 05), 20, ...) have a builtin MMU.  Therefore, this discussion is
>pointless.  The PDP-11 is a 256Kb-machine, *NOT* a 64Kb-machine!
>The 11/70 and all 22-bit Qbus machines have an addressing space of
>4Mbyte.  Again, with or without SepID.

True, but irrelevant.  A memory management context on the '11 is only 64Kbytes
(64K code + 64K data, if you have separate I&D), so unless you play funny
games with overlays (like they had to do with the V7m kernel, and like 2.9BSD
has to do to run large programs [like "vi"] on non-separate I&D machines),
you're still stuck with the 64K limit, no matter how much physical memory
your machine has.

Also, it's not necessarily true that all the "major" PDP-11's have built-in
MMUs (although, most useful configurations do include one).  For instance,
my old 11/24's MMU (as well as the one on my 11/23 - which is actually the
same CPU and uses the same MMU chip) was an optional feature, as was the
"Unibus map" (which allowed the 11/24 to access 22-bits worth of memory).

>> But other than that, it wasn't much different from regular V7.  Perhaps
>> that's what you had?
>_I_ had a vanilla 11/34 with 128Kw of memory running BSD2.9.
>The other machine ran V7 at first, but we put BSD2.9 on that too,
>since it was much faster.

V7 didn't run on the 11/34 (really - I have the original V7 documents
from Murray Hill that explicitly say this).  V7m would run on the 11/34,
because it had an overlaid kernel.  2.9BSD had an overlaid kernel, too,
like the V7m kernel.  As far as I know, no V7-based system that ran
on non-separate I&D machines ever came out of Murray Hill.

>Now, can we stop this part of the discussion, or move it to another
>group?  I think we are boring lots of other people....

Agreed.  I might suggest that anyone interested in continuing this
discussion do so on the INFO-PDP11 mailing list, which I maintain.
The mailing list is "info-pdp11@transarc.com"; requests to join and
other administrivia to "info-pdp11-request@transarc.com".

--Pat.