(click on photo to enlarge.
NOW WITH IMAGE MAP AND NAMES!!
I cut my teeth on DEC10's, first in high school, using the BOCES/NCODE KA10 (S/N 109) timesharing system on Long Island (via an ASR33 and an acoustic coupler), and then on KL1282 (running 6.03A and 7.01) as [1300,1523] and [301,273] at Stevens Tech, in lovely Hoboken, NJ. I was also a tourist on the MIT ITS systems as BUDD (a typo that stuck). From Stevens I took a job working on FORTRAN-10/20 at DEC Marlboro (using, KL1026/1042 as [31,5732], KL2137, MRFORT and others as <BUDNE>). When I left DEC, I managed the Boston University Computer Science 2060 (KL3047) as BUDD@BUCS20.ARPA at until it's demise in October 1986, when I became a full-time Unix person.
A great joy of using (and working on) DEC software was the focus on quality that simply doesn't exist in todays world. DEC's corporate culture was somewhat leaden, so the software wasn't always feature laden, but we took great pride in what we produced.
The flexibility (and hardware documentation) made the '10 a hotbed of research Operating Systems. BB&N created TENEX, and The major Computer Science Universities had their own Operating Systems: MIT had ITS, Stanford AI had WAITS, and CMU had a customized TOPS-10. All three later ran TOPS-20 systems, as did Rutgers, Utah and many others.
The first network e-mail was sent between two BB&N PDP-10's running TENEX.
The '10 was the center of Hacker culture, back when the word was high praise. PDP-10 assember was the Lingua Franca. The Jargon file, --MORE-- processing, EMACS, HAKMEM, "job control", and ZORK all originated from the MIT PDP-10 community. A great deal of software was available in source form, both from DECUS, the DEC users group, shared via mailed magtapes and on the ARPAnet. The PDP-6/10 community was preceeded by the PDP-1 community (which was preceeded by the TX-0 and TMRC communities)! Part I ("The True Hackers") of Steven Levy's book "Hackers: Heroes of the Computer Revolution" (the first two chapters of which are available via Project Gutenberg). Fred Hapgood's "Up the Infinite Corridor" has a chapter on TMRC. Another book to consult may be: Architects of the Information Society: Thirty-Five Years of the Laboratory for Computer Science at MIT by Simpson Garfinkel (MIT Press, 1999).
The PDP-6 was born in 1964, 1984 was the 20th Anniversary, the last major anniversary when 36-bit engineering was still going strong.
Somehow I scammed a copy of this booklet Twenty Years of 36-bit Computing with Digital 1964-1984.
At Alan Kotok's retirement, Alan and Rosanne autographed it for me (I briefly worked the code scheduler for VAST, the VAX to SAFE Translator). SAFE was a 64-bit VAX datatype compatible RISC project than Alan headed, before Cutler stole the project, renamed it PRISM along with all the instructions (so that they used PDP-11/VAX conventions to name the datatypes, so the native 64-bit load (LOAD64) became "load quadword", and they moved the dedicated zero register from R0 to R64). PRISM died, along with HL32, a 32-bit RISC project (named after the Hudson, Mass site at which it was being developed). However, it seems the developers of the Alpha AXP architecture were aware of PRISM. Some allege that the AXP stood for Almost eXactly Prism
Photograph by David Dyer-Bennet of the DEC Large Computer Group (LCG) in Marlboro, Mass (in front of building MR-1). (Photo taken for the 20th anniversary of 36-bit computing in 1984?) We don't know the exact date, but it was before June 23, 1984!
(click on photo to enlarge.
NOW WITH IMAGE MAP AND NAMES!!
New! DD-B found and scanned (with better resolution) the original slides which include a different shot from the same session, as well as seperate shots of smaller groups. Dan Murphy found photos he took of the shoot, including "behind the scenes" shots". Another 1984 photo of the TOPS-20 group (in suits!) which has a few folks who were away at the time of the above picture. There are more links to sites with photos in the documentation section of this page.
Peter Hurley's 1984 Life in the Fast ACs talk from 1984 DECUS.
Transcript of roundtable session hosted by Clive Dawson. Participants were: Gordon Bell, Robert Clements, Jim Fleming, Ralph Gorin, Richard Greenblatt, Alan Kotok, Dan Murphy, Glenn Ricart, Tony Wachs.
This is a copy of a page that used to hang in (later day TOPS-10 wizard) Jim Flemming's cube in MR1-2/L12. From The Book of the Weird, by Barbara Ninde Byfield, 1973, Dolphin Books, page 149.
|
Legend reads;
TRUST those Wizards who are thin and fine drawn. Mighty muscles, ruddy cheeks, and genial dispositions bespeak Wizards who when Apprentices and young men sneaked out to play instead of applying themselves to long days over hot crucibles and long nights pouring over scrolls, tomes, tablets and tracts. |
|
JMF's grave marker: TOPS-10 GURU. HAVE EDDT, WILL TRAVEL
Printed booklet of (latter day TOPS-10 Wizard) Tony Wach's struggle to adapt TOPS-10 to use the KL10's RH20 "internal channel" massbus controller. (can anyone tell me when this was printed, and for what occasion?)
|
|
Cover reads:
Tony in RH20 Land
|
I normally parked in back of MR-1 (DEC Marlboro, building 1), and walked in and out through manufacturing on the third floor (the site was sloped). Here are some stickers used on cabinets I grabbed:
cover |
instruction set diagram |
No DEC doumentation would be complete without this chart of the Powers of two and eight
All are works in progress; let me know of an corrections or additions!
My improved versions of programs for reading 9-track tapes (or tape images) written by DEC software on PDP-10's;
Other people's tape utilities;
Developed by PDP-6 hackers, who originally didn't want a timesharing system, ITS took an interesting approach to security; No operations were privileged (although many were certainly obscure). Sensitive operations were logged to the printing console. Guests accounts were available to "Tourists". Rather than make breaking the system a challenge, potential miscreants were encouraged to make postive contributions.
Mark Crispin wrote:
WAITS most advanced display support of all, not duplicated even today. Last shared code with TOPS-10 in 3.54 days, but was extensively hacked even then. Filesystem and device drivers were completely different (most reliable filesystem of any PDP-10 OS, the result of the world's worst hardware). Lots of funky devices.
Compuserve ran a highly modified system based on the PDP-10 4S72 Timesharing system. (ever notice that old CompuServe ID's were two 18 bit octal numbers?)
Tymshare's TYMCOM-X Operating System was based on the TOPS-10 Monitor, version 5.02C. It had some TOPS-20 features added, such as memory-mapped disk I/O with all disk structures based on a 512-word page instead of a 128-word block.
Many other funny tidbits can be found in the AI: HUMOR; directory, available as a gzip'ed tar, or live on SV-ITS
List of PDP-10 serial numbers (pdf)
The PDP-6 project started in early 1963, as a 24-bit machine. It grew to 36-bits to better support LISP. The PDP-6 was implemented using transistors on large "system modules". The price in 1964 of a PDP-6 system was $300,000, (CPU $120K) and could execute 250,000 instructions per second. Gordon Bell was the architect, Alan Kotok was the assistant logic designer. The PDP-6 designed to be used as a multiprocessor, and it's CPU had a seperate option number: Type 166. Gordon Bell wrote:
"The PDP-6 was designed for timeshared computation and real-time laboratory use with straightforward interfacing. At the initiation of the project, three timeshared computers were operational: a PDP-1 at Bolt Beranek, and Newman (BB&N) which used a high-speed drum that could swap an image in one 34ms revolution; an IBM 7090 system at MIT called CTSS, which provided each of 32 users a 32 Kword environment; and an AN/FSQ-32V at SDC which could serve 40 simultaneous users"
Design input for the PDP-6 came from MIT PDP-1 users including John McCarthy, Marvin Minsky and Peter Samson. Only 23 PDP-6's were built, and they were difficult to manufacture.
The KA10 (1967) CPU was a reimplementation of the asynchronous PDP-6 design using smaller modules and improved manufacturing. In order to convince management to make another shot at a large system, just about everything was optional: including fast memory for the accumulators, and the second set relocation/protection registers, floating point, and byte instructions. The minimum configuration, the 10/10 with 8K of memory and only paper tape (no DECtapes) was priced at $99,000. Oxford actually bought one of these (though they did have to buy more memory). Performance was improved to 380,000 instructions per second (1.5x the PDP-6) at a price of $150K for the CPU.
The KI10 (1972) CPU was implemented using integrated circuits. The CPU cost $200K and could execute 720,000 instructions per second (1.9x the KA) and had a synchronous clock, and a simple pager.
The KL10 (1975) was implemeted using 10K ECL and was microprogrammed. The CPU cost $250,000 and could execute 1.8 million instructions per second.
The KS10 (1978) was microprogrammed and implemented using TTL bit-slice.
A pair of PDP-10 (KA10) clones (prounounced "MAX") built at Xerox PARC. (Xerox refused to buy a PDP-10 from then competitor DEC), and instead offered an XDS Sigma 7; Alan Kay wrote:
"We decided it would take three years to do a good operating system for a Sigma 7, while we could build an entire PDP-10 in just one year."
See Chapter 7 "The Clone" in "Dealers of Lightning, XEROX PARC and the Dawn of the Computer Age" by Michael Hiltzik
MAXC had to be bug-for-bug compatible with the PDP-10's floating point instructions in order to run LISP (under TENEX). It used early iNTEL dynamic memory chips instead of core. The MAXC was somewhat faster than the KA10, and much more reliable (running more than 2000 hours between crashes more than once). Two were built, each operated for about 9 years.
After finishing the MAXC design, the PARC team started on the Alto, the prototype bitmap display & mouse driven workstation.
The PDP-6/10 were part of a lineage that started at MIT in the 1950's;
Project Whirlwind (operational in 1951) pioneered real-time interaction, the use of CRT displays and core memory. To test Whirlwind's core memory, a special system, the Memory Test Computer (MTC) was built by a team headed by Ken Olsen who had just graduated from MIT. Whirlwind engineers went on to develop computers at MIT Lincoln Laboratory in Lexington Massachusetts. Whirlwind was adopted USAF for use in the SAGE (Semi-Automatic Ground Environment) air defense system.
Other future DEC engineers who worked on Whirlwind and at Lincoln Labs, included Ben Gurley, Dick Best, Harlan Anderson, Stan Olsen, and Bob Savell.
The TX-0 was the first (zeroeth) transistorized computer, designed and built at MIT Lincoln Labs in 1956 to test transistor circuit designs and a 64KW, tube driven (S) memory for the TX-2 (The first computer designed, the TX-1 was rejected as too complicated, a step backward resulted in the TX-0). This echoed the development of the Whirlwind MTC.
TX-0 was designed by Wesley Clark (who would later design the LINC, which spawned LINCtape, which spawned DECtape -- DEC later sold the LINC, and integrated it with the PDP-8, first in the LINC-8, and later the PDP-12). TX-0 had an 18-bit word, a 5 MHz clock, and a 6 microsecond cycle time (it execute 83,000 additions per second). It used 3,500 Philco L-5122 fast surface barrier transistors (later commercially available as the 2N240) costing $80 each. TX-0 consumed 1000 watts, covered 200 square feet, and cost 3 million dollars to develop. Ken Olsen was in charge of constructing TX-0, and the display system (with light pen) was designed by Ben Gurley (who later designed the PDP-1). Memory design and fabrication were done by Richard (Dick) Best and Jack Mitchell.
The original TX-0 used 2-bits for order (op) codes and 16 for addressing. The four instructions were: Store memory (sto x), Add memory to accumulator (adx), Transfer on Negative Accumulator (trn x), the last was a microprogramming Operate command.
In July 1958 TX-0 was loaned to the MIT Research Laboratory of Electronics (RLE) with a transistor driven 4K word (T) memory. It lived in room 26-248, and was managed by Jack Dennis who extended the machine (by borrowing now unused address bits), increasing number of opcode bits from 2 to 5, doubling the memory to 8K words, and adding an index register. The TX-0 later inspired the design of the PDP-1. The TX-0 was used nearly 24 hours a day as a personal computer; by day it was used by researchers (Gordon Bell used it as a member of the Speech Research Group). Free time was used by "Hackers", many from the Tech Model Railroad Club (TMRC). Software implemented on TX-0 included word processing, sound recording, speech recognition, and music synthesis. Jack Dennis developed a macro assmbler named MACRO and an interactive symbolic debugger called FLIT (FLexowriter Interrogation Tape), named after a common household fly sprayed by a hand-pumped atomizer (or "Flit Gun"). Robert Saunders wrote the first MIDAS assembler.
Gordon Bell acquired the TX-0 as part of the DEC Computer Museum where is was operated November 13th, 1984 at a TX-0 reunion, afterwards it moved to the Boston Computer Museum, and is now at the Computer History Museum in California.
After working on the TX-0 and TX-2, Ken Olsen founded Digital Equipment Corporation (DEC) in 1957, which at first manufactured only "system modules" patterned on circuits first designed for the TX-0. While the company was founded with the explicit goal of producing computers, the company's backers felt the company should first establish a business building modules.
The PDP-1 (Programmed Data Processor) was DEC's first computer (but they didn't want to call it one), built in 1960. 50 were built and sold for $120,000 with 4K of 18 bit memory. The PDP-1 was designed by Ben Gurley using DEC's system modules, had a 5 microsecond cycle time. BB&N purchased the second machine built, a prototype, right off the floor of the 1959 Joint Computer Conference in Boston (The whole show was abuzz about this fledgling company and its little machine which cost less than $150,000). DEC gave an early system to MIT, where it was installed in a room next to the TX-0. MIT, BB&N and Stanford all implemented Time Sharing systems for their PDP-1's. (BB&N's was implemented by Sheldon Boilen under the supervision of John McCarthy).
At MIT Alan Kotok wrote the first DDT (named by Peter Samson?) for the PDP-1, a follow on the the TX-0 MicroFLIT debugger. Text processors included the Colossal Typewriter (by John McCarthy and Roland Silver), Expensive Typewriter (by Steve Piner, and extended by Peter Deutsch), the first TECO (written by Dan Murphy in 1962). The text formatter "Justify" by Peter Samson and was followed by TJ-2, (which all fed into the CTSS TYPSET and RUNOFF programs). The MIDAS assembler was ported to the PDP-1 over a weekend by a team of hackers. Peter Samson wrote a music compiler and player. Jack Dennis implemented hardware extensions to the PDP-1, and implemented a timesharing system for it.
Nearly half the PDP-1's built were used by ITT in ADX 7300 systems, for the ITT Telegraph Automated Telecommunications System.
The follow on to the PDP-1 was the PDP-4 (1962), with a goal of producing a smaller cheaper 18-bit system. Gordon Bell wrote "The PDP-4 took half the space of the PDP-1 and provided 5/8 the performance for 1/2 the price". The PDP-4 was not software compatible with the PDP-1 (but could use it's peripherals). The PDP-7 (1964) was conceived as a repackaged PDP-1, but since DEC had created a larger software library for the PDP-4, the PDP-7 implemented the PDP-4 architecture. The PDP-7 was followed by the PDP-9 (1966) and PDP-15 (1969). The majority of these systems were used for Process Control, Lboaratory, and Real Time applications.
The follow-on to TX-0, the 36-bit TX-2 became operational in 1958 (after Ken Olsen left). It was used to advance the state of the art in AI and man-machine interaction. It is most famous for the Sketchpad program by Ivan Sutherland which allowed a user to create and manipulate and resize sketches on the display by using a light pen. The TX-2 later ran timesharing, and in October 1965, the TX-2 exchanged the first packets with the AN/FSQ-32 at System Development Corp. in Santa Monica, California over a dedicated 1200bps line in an ARPA funded experiment. TX-2 was later on the ARPAnet.
Project Genie (1964-1969) at the University of California, Berkeley was funded by ARPA (contract SD-185) to produce a timesharing system. They started with the 24-bit SDS 930, and added modifications (which were later commercialized as the SDS 940). Several ARPA sites purchased SDS 940's, including BB&N, where it influenced the design of TENEX.
An offshoot of Project Genie was the Berkeley Computer Company (1968-1970), whose BCC 500 featured SDS 940 compatibility. Many BCC people were hired by Xerox'es Palo Alto Research Center (PARC), where the first computer they built was MAXC, a PDP-10 clone.
The AN/FSQ-32V was built by IBM for military purposes. It had a 48 bit word and 18 bit addresses. A spare system at System Development Corp. in Santa Monica, California ran an early timesharing system "TSS" funded by ARPA in 1963. The Q32 at SDC was never equipped with more than 48K words.
The SDC Q32 was networked to the TX-2
Jules I. Schwartz, E. G. Coffman, Clark Weissman, "A General Purpose Time-Sharing System", AFIPS Conf. Proc. (1964 SJCC), Sparton Books, Baltimore, pp. 397-411. (Best Paper award).
Coffman, E. G., Jr. Schwartz, J. I., and Weissman, C., ``Potentials in Large-Scale Time-Sharing Systems,''Proc. 2nd Congress, Information Systems and Sciences, Spartan Books, Baltimore, 1964.
Jules I. Schwartz, "Observations on Time-Shared Systems", Proc ACM 20th national conf, Cleveland Ohio, August 1965.
Jules I. Schwartz, Clark Weissman, "The SDC Time-Sharing System Revisited", Proceedings of the 1967 ACM 22nd national conference, pp. 263-271
Time-Sharing project started in September 1963 by Dartmouth professors John Kemeny and Thomas Kurtz. Ran on a Datanet-30 "Master" which served as front-end for communications, control, and scheduling, and a GE-235 back end. Began operations in May 1, 1964 at 4AM, by executing BASIC programs from two terminals simultaneously (both were "PRINT 2+2"). A fast one-pass ALGOL-60 compiler, a machine-language interpretive program DIP, and and editor EDIT were also available. Dartmouth Time Sharing System Reborn
phil at ultimate.com
This page has been viewed over 25,000 times
