23 Jan
2011
23 Jan
'11
10:32 p.m.
On 2011 Jan 23, at 9:11 PM, Evan Koblentz wrote:
Yes, I did see that and it sounds interesting, but the consensus for
the time being is still the Baby.
And one of the primary issues that led to the development of the stored
program concept. See below.
In the Universal Turing Machine sense, from computing theory. A UTM can
compute a certain class of functions. Any machine with a certain set of
capabilities can be shown to be a UTM equivalent and thereby able to
compute the same set of functions. With the assumption of being given
enough memory and enough time. That is, the assessment is made on
qualitative characterisation not quantitative.
All our general purpose machines today are UTM equivalents. Harvard
arch microcontrollers arguably excepted.
I think it would be difficult to construe the ENIAC into a UTM
equivalent.
Although this quote from the Wikipedia article is intriguing:
"Minsky goes on to demonstrate Turing equivalence of a counter
machine"
There was an interesting article a few years ago that showed the Zuse
Z3 (1941, harvard-arch relay machine) could be turned into a UTMeq by
providing it with a program implementing a stored-program pseudo
machine.
The point is that the EDVAC and the
stored-program-machine/UTM-equivalents that came after bore no
architectural resemblance to the ENIAC.
(Architecture here meaning the processor structure, not the
implementation (tubes, relays, etc.))
ENIAC's accomplishment was to achieve electronic speed computation. At
the same time it achieved that, it exposed it's own shortcoming: the
programming bottleneck. Enormous amounts of computing time were lost
while the program was changed. All those cables and plugs and switches
needed to be changed into electronic switches controlled by an equally
fast memory. Along with issues related to memory assignment, a radical
change in architecture was necessary and the stored program concept was
created. And there we step into the fog of what happened in those
meetings and how much Von Neumann with some knowledge of Turing's
Turing machine paper from the 30s influenced that creation.
Some interesting comments under stored program concept on:
http://en.wikipedia.org/wiki/Universal_Turing_machine
Keeping in mind the rhetoric.. if the ENIAC wasn't a computer then the
EDSAC gets that credit.
(I'd disagree about the 1948 stored program ENIAC, as I mentioned in a
previous message.)
Dismissing the
points by equating them with mechanical adding
machines is flippant. No, mechanical adding machines do not have the
properties mentioned, however mechanical machines with built-in and
multiplication and division were complex machines.
There are many complex calculating machines that aren't computers.
dismissing the ABC on the above is again
flippant. Multiplication and
division have been implemented as small programs throughout the
history of computers, be it in a hardware state machine, microcode,
or (frequently) as instruction-level programs as many machines do not
provide those ops as instructions. Various 'real computers' have
gotten by with little or nothing more than sign- and/or
zero-detection for data-sensitivity/conditional operation, just as
the ABC provides.
That m/d is one point among many; nobody's saying it was a useless
technique.
the Manchester Baby (first stored program
machine)
Not necessarily; see my email citing the 1947 EDVAC memo. I don't
like to overstate what the ABC did, which is why I prepared
that web article about it, to try to put down concisely what it did
technically. I think some have overstated what it did it on occasion,
but it is also about time the ENIAC supporters acknowledged what the
ABC did do, rather than just trying to dismiss it. I understand there
are the familial/emotional issues of having Mauchly's name dragged
through the mud and I can sympathise on that. And I understand there
may still be differing opinions on how much influence Mauchly
obtained from Atanasoff. But the technical assessments need to be
separated from those matters.
You're right. I am rewriting parts of the new web site to reflect this.
ENIAC .... took hours to change a program
That was certainly one of the biggest knocks against it. A fast
computer has limited utility if setting up the program was tedious. wasn't a
universal machine
Please elaborate. was an
architectural dead-end
That's not true, but I think you and I each made our points already
about ABC and ENIAC. was just an
overblown calculator, it simply wasn't a 'computer'
:-)
stored program machines, the Manchester Baby was
the first one to run
See above, re: Baby.
EDSAC was the first one to provide real service.
Oy! ENIAC was providing "real service", i.e. running important
applications for academia, government, and the Army, by 1945 (and with
a stored program by 1948). EDSAC wasn't even a pipe dream in 1945,
and didn't go into operation until 1949. Truce?
What? This list finally has a good thread about computing history and
you want to stop!? (Yes, I do have to get other things done, too.)