Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!vrdxhq!deller
From: deller@vrdxhq.UUCP (Steven Deller)
Newsgroups: net.arch
Subject: Re: 68000 Memory Managment (Bechtolsheim patent) (SUID Patent)
Message-ID: <2197@vrdxhq.UUCP>
Date: Fri, 10-Oct-86 02:33:02 EDT
Article-I.D.: vrdxhq.2197
Posted: Fri Oct 10 02:33:02 1986
Date-Received: Fri, 10-Oct-86 05:50:32 EDT
References: <727@sauron.UUCP> <610003@hpcnoe.UUCP>
Organization: Verdix Corporation, Chantilly, VA
Lines: 128
Summary: PDP-15 and PDP-12 were maligned, TOPS-10 NOT derived from PDP-15

In article <610003@hpcnoe.UUCP>, jason@hpcnoe.UUCP (Jason Zions) writes:
> Jerry Leichter says:
> 
> > I used TOPS-10 in 1969 or so; at the time it was already at Version 5 or
> > thereabouts.  JACCT was already there; I don't know exactly when it was
> > introduced.  TOPS-10 itself descended from earlier OS's for the PDP-15 (?),
> > which, for all I know, already had JACCT or some analogue.
> 
> TOPS-10 didn't really descend from any OS; it was sort of written from
> scratch with an interesting set of constraints. Someone from DEC wrote a book
> discussing the design of the early versions of TOPS-10; fascinating reading.
> 
> If TOPS-10 were to descend from any DEC OS, it would be the one that ran on
> the PDP-6, the predecessor to the PDP-10 (well, predecessor to the KA-10).
> If memory serves, the PDP-15 was a pair of -11's hooked up in tandem (in the
> same way a PDP-12 was a pair of PDP-8's in tandem). Exactly what the PDP-15
> and PDP-12 were used for, I couldn't tell you.
> 
> (Here I am, correcting a correction. If someone corrects me, we may set a
> record for levels of digression. This is net.arch, not net.ancient.history! )

Sorry to correct you, but your information is so far off base that it requires
correcting.  The PDP-15 and PDP-12 were not at all how you describe them.

The PDP-15 was an 18-bit computer, derived from the 18-bit PDP-9, which 
derived from the 18-bit PDP-7 (one of the original UNIX base machines), which
had a common ancestor with the 36-bit PDP-6 in the PDP-4.  The PDP-15
primarily added 8 index registers and slight modifications to the instruction
set to handle them, plus some better I/O controls.  The PDP-15 was the base
machine for RSX-15 (later RSX-15PLUS or just RSX-PLUS), which was the operating
system predecessor to RSX-11D (later IAS).  Hank Krejci (Hank are you out
there?) was one of the primary developers.  I programmed an Electrocardiogram
over-the-phone analysis system using it (including dialback to doctor's offices
and the like -- heady stuff for 1972).  The PDP-9 instruction set was a
glorified PDP-8 instruction set, i.e. hardly any instructions -- the first
RISC archtecture :-).

The PDP-12 was a 12-bit computer derived from the LINC-8.  The LINC-8 was a
join of the Laboratory INteractive Computer developed at MIT (I believe) and
the PDP-8.  The LINC instruction set had a number of innovative instructions,
particularly the "add two 12-bit integers plus the overflow carry" which
provided an easy method for efficient multiple-precision arithmetic.  It also
included a single instruction to read a tape block with or without wait
(can you say Linctape -- identical to DECtape except that DECtape wound from
the left reel to the right -- clever way around a patent :-) ), and several
single instructions for efficient CRT displaying of data.  I firmly believe the
LINC was one of the earliest innovators of application-relative CISC.  

The LINC-8 ran LINC code when decoding instructions in the 2nd of the eight 1K 
banks.  The PDP-12 allowed running either LINC or PDP-8 instructions anywhere;
each instruction set had one additional instruction that switched to the other
instruction decode mode.  Anyone in their right mind used the CISCish LINC set
for all real work.  The PDP-8 was only used for I/O and "preexisting" code.

This ability to switch the instruction mode anytime led to the (trivial) 
pursuit of trying to write machine code that did something useful when 
executed in PDP-8 mode, and something else useful when executed as LINC mode.  
This was not always so trivial; the base machine only had 8K 12-bit words.  
(Yea, we all KNEW there would be MB of memory to burn one day, but at the time 
8K of 12-bit words and a 32K word disk -- not 32M word -- was all you could 
afford for $20,000!!).  This trivial pursuit is, not suprisingly, very similar 
to the trivial pursuit seen in another news group of trying to write high-level
code that means "useful" things to two different languages; history always
repeats itself.

We had an multitasking system running on the PDP-12 that did real-time patient 
monitoring of 4 patients for ECG, cardiac output, blood pressure, temperature, 
plus several background jobs, including doing listings on the ASR-33.  Yes, it 
was a true multitasking system, and included an aggregate 2K A/D samples per 
second (with only 8K you had to process and dispose of the raw data real fast),
real-time annotated displays (it really helped algorithm development to see the
ECG waveform with cursor marks and numeric values, and to be able to freeze the
display any time), and "human-friendly" interactive diagnosis.  See the 1971 
Proceedings of the ACM Conference, p 682, if you really care (it doesn't hurt 
to do a little looking at history, just to give yourself some perspective).

Anyway, I always thought designers of instruction sets could learn something
from the PDP-12 (the PDP-11 instruction set seems to have learned some of its
instruction set there).  For the RISC proponents, there was a very nice
instruction set architecture proposed at U of Washington around that same time,
that provided for dropping into and out of (similar to PDP-8/LINC) a very 
effiecient, minimal instruction set, that was "optimum" for computation of 
expressions -- always thought that CISC and RISC might do better co-existing 
than either does alone -- particularly an appliction-related CISC and
expression evaluation RISC.  

Finally, for those of you not wise enough to have hit "n" before now :-), the
TOPS-10 system was derived from the PDP-6 OS with substantial influence from
Project MAC.  It in NO way derived from the PDP-15, which was released 2 years
after TOPS-10 had reached release 5.  The PDP-15 had DOS-15 (derived from
DOS/BATCH-9 and KMS-9) and RSX-15 as mentioned before.

I am also on shaky ground about the exact derivation of TOPS-10 - I just dug 
up a copy of DECSystem10 User's Handbook (yes, I'm a hell of a pack rat), but 
could not find any proper reference for the OS derivation.  However, it is 
true that the PDP-6 was the PDP-10 predecessor (both were 36-bit), and as said 
earlier, the PDP-4 was the predecessor to the PDP-6 (and the PDP-7).

Just in case you wonder about the other PDP numbers:
  PDP-1 was an 18-bit architecture that could be said to be the 
    predecessor of the PDP-4.  It used real old discrete logic B-series
    10MHz boards.  Hmmm -- 10MHz clock rate in 1962, and now we have, wonder
    of wonders, 16.67Mhz "high-speed" micros -- haven't come as far as some
    may think in 25 years. 
  PDP-2 and PDP-3 were aborted -- never finalized.
  PDP-5 was Edson DeCastro's first design -- it preceeded the PDP-8.  Edson
    went on to found Data General when his version of the "new instruction
    set architecture" lost out to Gordon Bell's PDP-11.  Sorry if I am
    slighting other designers -- they are the best known.
  PDP-13 was never born -- the number was excluded.
  PDP-14 was a non-computer, methodology for building discrete logic
    systems from R and S series flip-chips (DTL logic).
  
After the PDP-15, came VAX.

Sorry to be so long typed (as in long winded, nothing to do with 32-bit
integers), but I hate to see drastically incorrect information on the net.
And anyway, this has far more content and utility than either the discussion
on very large memory (as least the trivial digression on the number of
electrons in the universe), and the one on what average to use for benchmarks 
(if you cannot resist averaging benchmarks with no basis for weighting, only 
the geomtric mean is defensible -- we did not really need 20 articles stating
this simple fact :-) ).

Good night.

-- 
{verdix,seismo,umcp-cs}!vrdxhq!deller