16 Sep
2015
16 Sep
'15
3:18 p.m.
-----Original Message-----
From: cctalk [mailto:cctalk-bounces at classiccmp.org] On Behalf Of Noel
Chiappa
Sent: 16 September 2015 22:06
To: cctalk at classiccmp.org
Cc: jnc at mercury.lcs.mit.edu
Subject: Re: ENIAC programming Was: release dates of early microcomputer
operating systems, incl. Intel ISIS
a
From: Al Kossow
> the machine had to be configured (via
connecting up computing units
> with cables)
In 1947 ENIAC was modifed at BRL to be a stored
program computer.
Well, I did say "in the original ENIAC usage" it had to be configured by
plugging! I was aware of the later conversion.
Crispin Rope, "ENIAC as a Stored-Program Computer: A New Look at the
Old
Records", IEEE Annals of the History of Computing, Vol. 29 No. 4,
October 2007
Thanks for that pointer. I couldn't get access to that paper (it's behind
paypal I don't have the ability to pierce - I
would be grateful if someone
could
send me a copy), but in looking for it online,
I struggled to get it out of the University Library system.
I did find the very
similar:
Thomas Haigh, Mark Priestley, Crispin Rope, "Engineering 'The Miracle
of
the ENIAC'", IEEE Annals of the History of
Computing, Vol. 36, No.
2,
April-June 2014
which includes the same author, and is later, so hopefully more
definitive.
It's quite interesting: according to that, the conversion of ENIAC to a
'stored
program' configuration, after a period of about a
year of discussion and
planning, took place starting around March, 1948, and the first problem
was
run using it in April, 1948 - and it cites a lot of
contemporary documents
to
that effect.
(As the article points out, this contradicts the long-and-widely-held
impression, from a statement in Goldstine's book - and if anyone knew, it
should have been him! - that gave the date of that as September, 1948.)
Anyway, the new, earlier date is of course is very shortly before the Baby
ran
_its_ first program, in June, 1948. So there is a
rather interesting
question as
to which 'computer' ran first. I'd always
gathered it was the Baby, but
this
new data may overturn that.
I believe that its generally accepted that is true.
It is true that the 'program ENIAC' (to invent
a term to differentiate
that
stage of the machine from the earliest configurations,
which used the
cabling
method) did not store its program in the same
read-write memory as data,
as
the Baby did, instead storing it in 'EPROM'
(switches). However, I don't
consider that very important; nobody says that a machine running out of
PROM isn't a computer!
It is notable that in order to solve all problems, a computer must permit
self modifying code.
In the above article Crispin notes that ENIAC was succeeded by the IBM 701
and then omits the fact
that the 701 used the Williams tubes from the Baby as its main store, so
clearly both machines had a non-significant
influence on early computing. Without Williams and ENIAC there would have
been no IBM701...
The important thing is that it's a program, with things like subroutine
calls
from different locations, address modification for
data access, etc, and
the
'program ENIAC' apparently had all that (see
the list at the bottom of
page 51
in the article). So it's likely indeed be the
'first computer'.
Noel
What is and isn't a computer is always open to debate. The Manchester SSEM
or Baby is I would say the simplest thing that you could call a computer.
I could explain how to program the Baby to almost any one on this list. It
has seven instructions, so in a 32 bit word, a 5 bit address and a 3 bit op
code.
Crispin Rope concentrates on the power of ENIAC and its usefulness, neither
of which can be argued with, but to me a "computer" without self-modifying
code is a programmable calculator even if it has index registers...
If you want to rate a computer by the work it did ENIAC was far more useful,
but the SSEM was never intended to be a "Useful" computer. It was a testbed
for the Williams Tube.
... on the other hand whilst both added to our knowledge of computing in the
longer term neither were IMHO especially influential going forward.
Self-Modifying code became the norm, and Williams Tubes were rapidly
superseded by core...
From what I have seen the UK was slow to move onto Core
store, probably
because IBM bought the patents, so whilst the IBM 704 was already
using Core
in 1954 the Ferranti Pegasus from 1956 went back to using Nickle Delay Lines
and a Drum for "Main Store" although IMHO the Delay Lines were really main
store, from what I remember all code must be in the delay lines. The LEO
(Lyons Electronic Office) also used Mercury Delay Lines....
Dave Wade