I just finished reading
"Why Computers Are Computers: The SWAC and the PC"
by David Rutland
Wren Publishers
PO Box 1084
Philomath, OR 97370 USA
ISBN 1-885391-0506 (Hardcover)
1-885391-06-4 (Softcover)
Copyright 1995
The author tried to write two books at once. He really wanted
to explain to the novice that the modern PC is essentially the
same as the SWAC, built in 1950. By this, Rutland means that
both are "stored program machines", and I skipped over sections
where he tried to make this point. But I did learn that the
credit does not all go to John von Neumann!! That was a huge
surprise to me.
Of course, I was really interested in Mersenne primes since the
SWAC was the first computer to find new Mersennes, 5 of them in
all were found by Robinson. Also, Alex Hurwitz cut his teeth
on the SWAC, although he found M(4253) and M(4423) on an IBM 7090.
The SWAC was dedicated on August 17, 1950. Regarding
primes, Rutland writes on pages 126-127:
"Some months after the dedication, the SWAC was running
well enough for a program to be run for hours
without error. The mathematicians at INA, like
mathematicians throughout the world, were interested
in prime numbers. Numbers like 3, 5, 7, 11, 13 and
so on that cannot be divided exactly by any number
but themselves. As the numbers get bigger the primes
keep getting further and further apart and are more
difficult to find. Each number must be tested to see
if a smaller number will divide into it exactly.
Mathematicians who deal with the theory of numbers
are very interested in finding large ones, so they
programmed the SWAC to start with a known large prime
and find a larger one. It was set to this task and
after hours of calculating, it came up with a prime
number larger than any yet known. This was an
achievement only of real interest to mathematicians.
But to all of us that had built a computer from
'scratch' in less than 18 months, it was like a
milestone in the history of computers."
Attention Chris Caldwell! Reference Table 2 on
http://www.utm.edu/research/primes/notes/by_year.html
Is this a previous record? Rutland writes that a record
was set in 1950. All I have seen was Miller and Wheeler
setting records on EDSAC on Jun 7, 1951:
http://www.scruznet.com/~luke/lit/lit_062s.htm
http://www.scruznet.com/~luke/lit/lit_063.txt
Aside, didn't they find 934(2^127-1)+1 and then 978(2^127-1)+1
before 180(2^127-1 )^2+1 ?
Regardless, Rutland's only mention of Mersennes is on page 143:
"The mathematicians continued to look for high
prime numbers. The particular primes that the SWAC
looked for are called Mersenne numbers, which are
calculated by the formula 2^p-1. That is, 2 raised
to the p power minus 1 in which p is a known prime
number. For example, if p equals 3, then the
Mersenne number is 2 cubed (2^3), or 8 minus 1
which equals 7, a prime. But all Mersenne numbers
are not prime and must be tested to see if they
can be divided by a different number than themselves
and 1. The first few prime Mersenne numbers, when p
is equal to 2, 3, 4 and 7, are 3, 7, 31 and 127. As
p gets larger, the numbers get bigger very quickly and
testing to see if they are prime also gets longer.
After 453 hours, SWAC found a prime with p equal to
2281, which when written down would be over 700 decimal
places long!
M2281 was the 5th and last Mp that Robinson found. The figure
of 453 hours is probably cumulative, and not just to test
M2281 only. Robinson announced his 5 Mersenne Primes here:
http://www.scruznet.com/~luke/lit/lit_024.txt
the first two were found on January 30, 1952. He also wrote
that it took about 5 minutes per small p and one hour for
each large p.
There a wealth of information in the book. Rutland was a young
Electrical Engineer (Cal Tech) who helped build the SWAC. He
describes it in detail and compares it to other machines of
the day. Lots of great photographs. I apprectiated his
description of early memories, especially Mercury delay lines
and Willams tubes. Amazing! Perhaps the best of all was the
shift register: just looped cables, "electrical delay lines" :-)
Rutley writes they programed in hexadecimal 0 through 9 and A
through F. But Robinson,
http://www.scruznet.com/~luke/lit/lit_024.txt
used 0 through 9 and u, v, w, x, y, z.
I'm still confused about the SWAC word size. Rutley often mentions
37 bits. But then he writes
"These lamps were arranged in eight groups of four for the
36 bits of the SWAC word, plus an extra lamp for the 37th
bit which represented the sign of the number."
Maybe each "byte" had a parity bit? I doubt it because the
arithmetic unit was 37 bits wide. Are you reading this, Alex?
Here's something that hasn't changed in 50 years:
"When they found an error, they would of course
report it to the engineers, who generally tried
to blame the programmer by saying 'maybe the
program is wrong.'"
I'm *still* hearing that. I'm telling you, it is NOT
my code! It the guard's two-way radio interfering
with the USB and I can prove it!! 3 timeouts cause
a reset and it's in the spec!!!
Amazon.com
US$24.95
3-5 weeks estimated delivery
http://www.amazon.com/exec/obidos/ASIN/1885391056/qid=943260411/102-6052119-6422430
I recommend this book.
--Luke
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
