Subject: Re: CTIX architecture
To: John E. Clark <firstname.lastname@example.org>
From: Mike O'Brien <email@example.com>
Date: 12/09/1994 12:49:04
At 10:51 AM 12/8/94 -0800, "I presume I need no introduction."
>Didn't the *very* first release of BSD swap full images as well as do
>paging? I thought that BSD was capable of complete swaps as far forward
>as 4.1 (but my memory may be failing me -- I was a newbie just at the
>release point of 4.2. Man, talk about a slow system!).
Then John Clark said:
>I was thinking about VAX'es as a target for which a thing called un*x V-32
>was for. But it is true that Sun-1,2's were running some form of BSD like
I knew there was a reason for getting old. It's so your memory turns
out to be worth something.
The first version of UNIX turned out by AT&T, Unix 32/V, was indeed a
32-bit port of more-or-less straight-up PDP-11 UNIX that ran on a VAX in
swapping mode. It worked fine. We ran it at RAND for about six
months, waiting for BSD UNIX to come out. Once Ozalp Babaoglu did
the VAX paging code there was no looking back.
Of course things were pretty strange in those days. Setting a bit in the
processor status word on a VAX makes it suddenly start interpreting PDP-11
instructions. UNIX support was built in for this so that Ken Thompson's
PDP-11 binary-only chess program would still work. But it actually
provided an entire PDP-11 file system and everything, so you could run all
your old (non-floating-point) PDP-11 programs. We ran PDP-11 'nroff' just
for grins. It ran faster than VAX 'nroff'. By a lot. We eventually
figured out that this was because it was using 16-bit integers and
instructions, and the VAX version was using 32-bit integers and
variable-length instructions, and since the memory system ran in
quadwords, PDP-11 'nroff' was just sort of romping through the machine
at about twice the speed. It was a sobering thought.
Not that this matters diddly poop in a 386/486/585.999998265 world.
But there you are.