24 Jan
1999
24 Jan
'99
5:58 a.m.
<On Sat, 23 Jan 1999, Allison J Parent wrote:
<> Unix didn't always have the idea of virtual memory. See Minix for an
<> example of that also V4 and maybe V5 unix didn't either.
<
<I would expect that the original PDP7 that ran UNIX didn't have VM...
No it didn't. I'm fairly certain that VM didn't appear until V7 on pdp-11
or maybe later.
The fact of the matter is there have been 8bit unix impmentations. The idea
that there is only 64KB is just means more reliance on things like overlays
or paging/swapping. Also many of the mid 80s 6809 and z80 system had some
kind of MMU to support large address spaces. The Z180 was good for 1mb with
it's MMU and most of the paging schemes for s100 permitted anyware from 512k
to 16mb. the late z280 allowed 16mb space using a paged MMU and added I&D
space along with user/system spaces. The latter I&D space (instruction and
data spaces) were features only found on the upper models of PDP-11
(11/44,11/45,11/60 and 11/70 and the J11 chip based systems 11/73 and
11/83). With I&D space and user/system space VM was supportable.
Allison
24 Jan
24 Jan
7:41 a.m.
New subject: VM on PDP-11 (was Re: Unix for 8080/Z80? [Re: WooHoo!! PC/XT Unix anyone?])
Allison J Parent <allisonp(a)world.std.com> wrote:
No it didn't. I'm fairly certain that VM
didn't appear until V7 on pdp-11
or maybe later.
I'm fairly certain that VM didn't appear in ANY released version of Unix
on a PDP-11.
Only a few models of PDP-11 had the processor capabilites needed to support
VM (instruction continuation or instruction restart), and they were not used
in any commercial OS. It's been rumored that there were bugs that rendered
the capability unusable.
In any case, though, the PDP-11 only supported 8 pages of 8K bytes (or 8 pages
each of I & D space), which isn't really enough to make VM noticably superior
to swapping. A PDP-11 with 4M of physical memory could support around 30
maximum-sized processes resident in memory simultaneously.
8:45 a.m.
No it didn't. I'm fairly certain that VM
didn't appear until V7 on pdp-11
or maybe later.
IIRC, adding virtual memory support to UNIX was the project goal for the
CSRG at Berkeley. That was what BSD Unix was all about, and it was why the
kernel image was /vmunix vs. the /unix which was the prior kernel image name.
--Chuck
8:58 a.m.
New subject: PDP-8 prices
Sigh. The "PDP-8 in unknown condition" sold for $1526 on Epay. Wonder what
a cherry 8/E with all the trimmings would fetch? Ugh.
--Chuck
10:46 a.m.
New subject: PDP-8 prices
A non-text attachment was scrubbed...
Name: not available
Type: text
Size: 626 bytes
Desc: not available
Url :
http://www.classiccmp.org/pipermail/cctalk/attachments/19990125/04c00167/at…
12:06 p.m.
New subject: Unix and VM
No it
didn't. I'm fairly certain that VM didn't appear until V7 on pdp-11
or maybe later.
IIRC, adding virtual memory support to UNIX was the project goal for the
CSRG at Berkeley. That was what BSD Unix was all about, and it was why the
kernel image was /vmunix vs. the /unix which was the prior kernel image name.
--Chuck 
12:02 p.m.
In message <199901241958.AA15845(a)world.std.com>om>, Allison J Parent writes:
<On Sat, 23 Jan 1999, Allison J Parent wrote:
<> Unix didn't always have the idea of virtual memory. See Minix for an
<> example of that also V4 and maybe V5 unix didn't either.
<
<I would expect that the original PDP7 that ran UNIX didn't have VM...
No it didn't. I'm fairly certain that VM didn't appear until V7 on pdp-11
or maybe later.
There've already been a few followups on this and I'm to lazy to mix
all the references together. So I arbitrarily pick this point in
the thread to jump in.
The complete answer depends on what one means by VM. (Hey, here we
go with the definitions again.) If you take the pure viewpoint
and say that VM is the distinction between a virtual address and
a physical address, then all you need is an address translation
unit. These memory management units (MMUs) on PDP-11s were used
pretty early on. I don't know exactly how early, however. The
first 11 UNIX was ported to was the 11/20 (only called that after
other models were introduced; the 11/20 was the first model of the
11). If the story I heard in college is correct, there were only
three examples of a prototype MMU ever built for the 11/20 and
the one I saw had as many cards as the CPU. I doubt that Bell
Labs worked with one of those. The next model they ported to
was the 11/45 which had a rather good MMU including separate I/D
spaces. I don't know where along the line MMU support was added,
but it was definitely there by the time 6th edition came along
and John Lions wrote his masterpiece on it. One of the tables
in Peter Saulus's book suggests that it must have happened
somewhere between 3rd edition and 6th edition.
If by VM one refers to the ability to swap individual pages and
to do demand paging, then there were several attempts to
get paging into UNIX. The first was purely experimental using
an 11/55 (or 11/60, I can't remember). I don't remember my
source, but I do remember it reporting that performance was
very poor, and the line of research was not followed any
further. When Berkeley took 32V (the Bell Labs VAX port)
and added paging, they released 3BSD as Chuck mentioned.
As usual, if you want to look further at UNIX history, look
at the early papers and at Salus's book *A Quarter Centry of
UNIX*.
Brian L. Stuart
9465
days inactive
9465
days old
6 comments
6 participants
participants (6)
-
allisonp＠world.std.com -
ard＠p850ug1.demon.co.uk -
blstuart＠bellsouth.net -
cmcmanis＠freegate.com -
eric＠brouhaha.com -
pechter＠monmouth.com