Totally new subject...base 10 digital computers
Joseph Crowe
I have always wondered this. Has anybody ever architected and
built a digital computer based on base 10 (decinary?) rather than
base 2 (binary) logic? I can just imagine how fun it would be to have
dit (decinary digits) flags that could represent false, true and eight
levels of maybe. If such a machine were possible would various
voltage levels represent different values? Inquiring minds want to
wander!
Joe
Marty Itzkowitz
numbers as strings of characters. ADD added the string at the B address to
the string at the A address, and left the result as a string at the B address.
The length of the field was given by a word mark in the B string. Overflow was
handled by setting Zone bits in the B string high-order position.
The IBM 1620 was decimal, and did its arithmetic by table lookup. Very strange
things could be done by altering the contents of these tables.
Both machines date from the late 50's and early 60's.
Marty Itzkowitz
John Mashey
>The IBM 1401/1460 series were character oriented machines that represent
IBM 7070 -> 7074 (early-mid 60s) were decimal machines;
no octal or hex core dumps. [7074 was my first machine, running a local
Penn-State developed OS, and a FORTRAN-like thing called DAFT - Dual
Autocode-FORTRAN Translater. Nice software for its time, symbolic
core dumps, etc. Something like 10K words of memory [5 digits each?
10 digits each? something like that; it's been a long time; people were
less than pleased to go to a 360/50, despite much better technology.]
They actually got used for a long time thereafter;
one story is that in the 70s, PacTel actually bought every one it could
for replacement parts to keep an important application running.
[Converting call-acitvity tapes into bills!].
--
-john mashey DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: ma...@mips.com OR {ames,decwrl,prls,pyramid}!mips!mash
DDD: 408-524-7015, 524-8253 or (main number) 408-720-1700
USPS: MIPS Computer Systems M/S 5-03, 950 DeGuigne, Sunnyvale, CA 94086-3650
kai...@heron.enet.dec.com
The very first machine I ever programmed, the Burroughs/Datatron 205, was a
(vacuum-tube) machine in which each addressable word consisted 10 decimal
digits. 1961.
___Pete
kai...@heron.enet.dec.com
+33 92.95.62.97
Douglas W. Jones,201H MLH,3193350740,3193382879
> Hi Folks
> I have always wondered this. Has anybody ever architected and
> built a digital computer based on base 10 (decinary?) rather than
> base 2 (binary) logic?
BCD machines were once a dime a dozen. I've got a collection of old
manuals sitting around, and it's amusing to see what people did with
the extra values that you could put into the 4 bits that the ALU understood
as the representation of one decimal digit. Even the intel 8080 had
pretty some support for BCD arithmetic, and machines like the VAX and
the IBM 360 (etc.) have a complete set of BCD instructions.
I don't know if anyone has seriously pursued the idea of building a
decimal machine using actual decinary logic, with 10 voltage levels
per single wire conveying one digit. On the other hand, if you read
the old science fiction story "When HARLIE Was One," they (the people
in the book) credit HARLIE's intelligence to the use of 12 valued
internal logic instead of binary logic. The truth is, two valued
logic is always sufficient, in the sense that you can use it to
approximate any other logical system to an arbitrary precision if
you use enough bits in a word (for 12 valued logic, 4 bits allows
an exact approximation, for example). I think the author was actually
thinking in terms of something like fuzzy logic, but that doesn't
require any new hardware invention, and a new number base offers
no help at all to the fuzzy folks.
Doug Jones
jo...@cs.uiowa.edu
Keith Bierman fpgroup
Many handheld computers are decimal.
There is an IEEE standard for decimal fp (854).
Not all bases are equally good; for floating point binary has the
benefit of minimal wobble. The paper by Goldberg "what every scientist
should know about computer arithmetic" has a good discussion of this.
--
----------------------------------------------------------------
Keith H. Bierman keith....@Sun.COM| k...@chiba.Eng.Sun.COM
SunPro 2550 Garcia MTV 12-33 | (415 336 2648)
Mountain View, CA 94043 <speaking for myself, not Sun*> Copyright 1992
Robert Hyatt
Yeah. A couple come to mind (although they sure date back a while and I
might have dropped a few bits over the years.) The IBM 1620, the RCA
301/3301 are but two. Think this dated to the old use of packed decimal
arithmetic as opposed to floating point and or fixed point binary which
they didn't have.
--
!Robert Hyatt Computer and Information Sciences !
!hy...@cis.uab.edu University of Alabama at Birmingham !
Tim McCaffrey
decimal, right down to addresses. In fact, it is difficult to
do binary arithmetic on them. The primary programming language
is COBOL, and the instruction set shows it. (C is pretty much
impossible to implement on these machines).
Tim McCaffrey
Michael O'Dell
the 1620 was a decimal machine with a variable "word length".
futher, it did sums via table lookup in tables stored in memory
(Model I - the Model II had an adder but still did table-driven
multiplies), so you could even change the base (downward) if you
were very, very careful. it was not exactly spritely (20usec for
a cycle, and more than a few cycles/instruction, digit serial).
you could compose small programs sitting at the console typewriter
with considerable ease. the most interesting part was the you
could write Business programs in Fortran and never worry about having
to round to cents because of the binary inadequacies.
it was a real shocker to run on a binary floating point machine
for the first time.
it was a truly marvelous machine - my first computer.
"COLOSSUS - The Forbin Project" showed the largest collection of
1620 front panels ever assembled anywhere. now there was a machine
with a REAL front panel.
anyone have one of those panels around interfaced to a SPARCstation????
-Mike O'Dell
Richard Kenner
>on base 10 (decinary?) rather than base 2 (binary) logic?
From a footnote in "Microprocessors: A Programmer's View" by Robert Dewar
and Matthew Smosna:
"The first author's uncle worked for Plessy's (a large computer firm
in England) at one time and was involved with their very first
computer. This machine (called the PEP) had registers for
representing numeric quantities consisting of a row of decimal
devices, followed by a binary device, a decimal device, and a 12-state
device (it was called a duo-decatron). Even the British might have
forgotten that that is a reasonable format for pounds, shillings (which
went up to 20) and pence (which went up to 12), because the British
long ago changed to a decimal money system. This is a remarkable case
of hardware that *really* knew what its domain was going to be!"
Bob Bentley
(Michael O'Dell) writes:
>
>it was a truly marvelous machine - my first computer.
>
Mine too - a 1620 model IID ("D" standing for disk-based). Its operands
were variable length, which made it great for solving certain classes of
problem - like whether 196 ever became palindromic!
Who else remembers what the significance of "4900796" was? Followups
should probably go to alt.folklore.computers.
--
Bob Bentley
Intel Corp., M/S JF1-19 E-mail: r...@ichips.intel.com
5200 N.E. Elam Young Parkway Phone: (503) 696-4728
Hillsboro, Oregon 97124 Fax: (503) 696-4515
Joseph Crowe
When I posted this question I should have made it clear that what I
really meant excludes Binary Coded Decimal (BCD) which was popular on
many 50's and 60's vintage computers. What I really meant was true
10-state discrete logic. That is, a single dit (decinary digit) could
have values from 0-9 and a 32 dit word could have 10**32 distinct
values. The addressing range would be enormus, but not enough to
address every partical in the universeB^).
J.
John R. Levine
>What I really meant was true 10-state discrete logic.
Ah, that's slightly more exotic. The Eniac and the Harvard/IBM Mark
I were really decimal. The Eniac used 1 out of 10 logic with 10 flip
flops per digit, the Mark I used some combination of relays and
rotating cams.
I doubt anybody ever got a 10-level logic to work, the most I ever heard of
is 4 level in the 8087.
Regards,
John Levine, jo...@iecc.cambridge.ma.us, {spdcc|ima|world}!iecc!johnl
Jim Haynes
>really meant excludes Binary Coded Decimal (BCD) which was popular on
>many 50's and 60's vintage computers. What I really meant was true
>10-state discrete logic.
Then I think the history begins and ends with ENIAC, where the logic was
10-state, but in the form of serial pulses rather than 10 discrete voltage
levels.
--
hay...@cats.ucsc.edu
hay...@cats.bitnet
"Any clod can have the facts, but having opinions is an Art."
Charles McCabe, San Francisco Chronicle
Hans J. Mayer
[...]
|> ... I can just imagine how fun it would be to have
|> dit (decinary digits) flags that could represent false, true and eight
|> levels of maybe. If such a machine were possible would various
|> voltage levels represent different values? Inquiring minds want to
|> wander!
Isn't that what fuzzy logic claims to be? :-/
I think next year I'll try to invent an analog computer :-)
--
Hans J. Mayer, hma...@venus.darmstadt.gmd.de (or ma...@gmdzi.gmd.de)
German National Research Center for Computer Science (GMD)
Anthony Shipman
>Hi Folks,
The very first computer design, the Analytical Engine by Babbage ca 1830,
was base 10. A "dit" was a gear wheel with one of ten possible positions.
The memory store used 50 dit words. (All integer, no floating point!)
Ah but if he had thought of making it a binary machine he might have managed
to build one.
--
Anthony Shipman "You've got to be taught before it's too late,
C. P. Software Before you are six or seven or eight,
19 Cato St., East Hawthorn, To hate all the people your relatives hate,
Melbourne, Australia, 3121 You've got to be carefully taught." R&H
Donald Lewine
Yes. I learned to program on one in 1960: The IBM 1401. The machine
I used had 4000 (not 4096) positions of core. It also was a variable
wordlength machine so a number was as many digits as you wanted.
--------------------------------------------------------------------
Donald A. Lewine (508) 870-9008 Voice
Data General Corporation (508) 366-0750 FAX
4400 Computer Drive. MS D112A
Westboro, MA 01580 U.S.A.
uucp: uunet!dg!lewine Internet: lew...@cheshirecat.webo.dg.com
Malcolm Shute
>really meant excludes Binary Coded Decimal (BCD) which was popular on
>many 50's and 60's vintage computers. What I really meant was true
>10-state discrete logic.
What interested me, though, was your next bit:
>That is, a single dit (decinary digit) could
>have values from 0-9 and a 32 dit word could have 10**32 distinct
>values. The addressing range would be enormus, but not enough to
>address every partical in the universeB^).
a binary computer with more than 10**32 addresses.
In general, a base-n flip-flop needs n.k transistors.
(In discrete logic, k=(1 transistor+1 resistor), in MOS
it is perhaps k=2).
If you want to build a computer with the capacity to represent
some enormous number, X, then you are going to need log[n](X)
digits in your base-n system, and hence n.k.log[n](X) transistors
per register (and register equivalent --- adders, subtractors, etc),
and thus K.k.log[n](X) transistors overall (where K is the number
of registers and their equivalent).
Amount of hardware = H = K.k.n.log[n](X) = K.k.n.ln(X)/ln(n)
dH/dn = K.k.ln(X).( ( ln(n) - 1 ) / ( ln(n)^2 ) )
dH/dn = 0 when: ln(n) = 1, ie. when n=e (2.71828..)
Thus, the least hardware would be involved in your computer for storing
this arbirary number, X, if you built it using the base-2.71828 numbering
system.
From this, it turns out that a base-10 computer is going to be more
expensive (less efficient use of its hardware) than a base-2 computer.
Notice, though, that a base-3 computer might be better than a base-2 one.
Some research groups are seriously toying with ternary logic.
It does indeed offer a number of advantages over binary... but has
been prevented from coming into mainstream use by a number of disadvantages
(which my K.k.n.log[n](X) expression ignored).
(BTW isn't the base-10 system more generally called "decimal", or is there
a reason for you using this other name?)
--
Malcolm SHUTE. (The AM Mollusc: v_@_ ) Disclaimer: all
Joel Berman
There was some work done in the 60's on trinary machines. I worked on one of
them which had +5, 0, and -5 volts for logic levels. We were trying to solve
the problem of reliability of backplane wiring. With todays problems of running
out of wires before running out of transistors on a chip, we may see a
reemergence. I'll have to see if I can remember the truth table for a trinary
`or' gate.
/jb
peter da silva
since each piece of metal could now carry more information per transition?
--
-- Peter da Silva, Ferranti International Controls Corporation
-- Sugar Land, TX 77487-5012; +1 713 274 5180
-- "Have you hugged your wolf today?"
Malcolm Shute
>The very first computer design, the Analytical Engine by Babbage ca 1830,
>he might have managed to build one.
In fact, of course, the Analytical Engine used a mixture of base-10 and base-2,
depending on convenience to the hardware which Babbage was designing.
jeff bell
>A base 10 computer would use more transistors, but wouldn't it use less metal
>since each piece of metal could now carry more information per transition?
Yeah, but you now have to swing over 9 times as much noise margin.
If you cut noise margins, then you have to add shielding somehow.
Your settling time will no have to wait until any overshoot is past.
You now have to control your process enough to have 9 threshold voltages
come out right.
--
-Jeff Bell "I guess that's why they call it code."
jb...@danger.enet.dec.dom
Dean McCullough
a character, and thus could be considered a 6 bit byte binary computer.
Though many have mentioned the 1620, I am surprised that no one has
mentioned the earlier machine, the IBM 650. This used a system called
bi-quinary, where a digit was represented by two pulses. The time between
these (five possible values) and the relative direction of the second
pulse to the first provided the ten possible values.
Don Wells
dpm...@afterlife.ncsc.mil (Dean McCullough) writes:
DM> ... Though many have mentioned the 1620, I am surprised that no
DM> one has mentioned the earlier machine, the IBM 650. This used a
DM> system called bi-quinary...
Yes, the IBM 650 was decimal, using bi-quinary logic. One of its
principal competitors was the Burroughs Electrodata 204; it too was
decimal, but my memory is that it used BCD logic, i.e. hex [4-bits]
but only using 10 of the 16 states. I coded many machine language
programs for the 204, and thought of it as a decimal machine.
Both of these machines had *wonderful* front panels, with the bits of
the registers displayed in banks of little neon bulbs. The machines
were slow enough -- varying from about 50 to about 2000 instructions
per second depending on drum memory rotational latencies -- that you
could often see interesting patterns in the lights.
--
Donald C. Wells Associate Scientist dwe...@nrao.edu
National Radio Astronomy Observatory +1-804-296-0277
520 Edgemont Road Fax= +1-804-296-0278
Charlottesville, Virginia 22903-2475 USA 78:31.1W, 38:02.2N
Sarr J. Blumson
That was the 1401/1410. The 1620 needed two decimal digits, each encoded
in four bits, to store a character. There was a fifth bit, which was
sometimes the sign and sometimes told you where the word ended.
(Sarr Blumson)
Dick Smith
>>... a base 10 machine, 6 bits to represent
>>a character, and thus could be considered a 6 bit byte binary computer.
sa...@sinshan.citi.umich.edu (Sarr J. Blumson) writes:
>That was the 1401/1410. The 1620 needed two decimal digits, each encoded
>in four bits, to store a character. There was a fifth bit, which was
>sometimes the sign and sometimes told you where the word ended.
Actually, the 1410 had a 8-bit-wide memory: 6 bit characters, plus a 7th
bit called "word mark" which indicated where a multicharacter word
ended, plus a parity bit. If I remember correctly, the arithmetic unit
was all built with 2-out-of-5 logic with error checking at every step.
A 100-digit divide would execute in about 10 seconds... it was nifty!
--
Dick Smith smi...@ast.dsd.northrop.com
Sarr J. Blumson
Where overflow is loosely defined, since both machines were variable
"word" length: the A and B addresses were the addresses of the LOW order
digits, and they moved to the left just like a human till they found
digits with the zone bit set. In other words, infinite (up to the size
of memory) precision arithmetic in the hardware (to the extent that
arithmetic was in the hardware at all in the 1620, anyway).
sarr
(Sarr Blumson)
Ken Lerman
| I have always wondered this. Has anybody ever architected and
|built a digital computer based on base 10 (decinary?) rather than
|base 2 (binary) logic? I can just imagine how fun it would be to have
|dit (decinary digits) flags that could represent false, true and eight
|levels of maybe. If such a machine were possible would various
|voltage levels represent different values? Inquiring minds want to
|wander!
|Joe
The IBM 650 had a word size of 10 digits. Main memory was 2000 words
on drum. To the user, it appeared to be a decimal machine. If you
looked at the display, you got the impression that it was a bi-quinary
machine. (Does anyone out there know if it was really bi-quinary?)
The IBM 1620 was also a decimal machine. I believe it was available
with 20K, 40K, or 60K digits of memory. It was a variable word length
machine which permitted such (perverse) actions as erasing all of
memory with a single instruction (or bug).
The 1620 used binary (two voltage state) logic internally. I've
always assumed the 650 did, too, but I have no real evidence.
Ken
--
ler...@stepstone.com (203) 426-1875
The Stepstone Corporation (203) 270-0106 (FAX)
75 Glen Road
Sandy Hook, CT 06482