Computers I have used

The following list names most of the computers and operating systems I have used, during the rather long time (over 45 years) since I started using computers! Many of the items in the list include links to further information. In some cases, I have used the same hardware on more than one operating system, or even standalone (no operating system). The {a,b} notation means an expansion of the text to include all of the comma separated items, e.g. HAL-{8000,9000} means HAL-8000 and HAL-9000.

This is mostly in chronological order. I've tried to keep each entry self contained, so there is some repetition.

ICL 4130 under NICE and KOS

University of Kent, 1970-1976.
Originally manufactured by NCR Elliott, as the Elliott 4130 (part of the 4100 series). A 24 bit, word oriented machine.
Input was via punched cards and paper tape, output to 300 lines-per-minute line printer or punched paper tape. Online system was the Kent Online System (KOS) via 8 teletypes (6 housed under the stairs in the Cornwallis foyer, now a cleaners' cupboard). Storage was on ½ inch 800bpi magnetic tape, and four 1MB exchangeable disk drives. We later upgraded those to 2MB drives!
I learned the assembly language on this machine (the first assembler programming I'd done); the language was called NEAT. My first program did simple linear regression on data points. I also programmed in BASIC, ALGOL-60, FORTRAN IV, and BCPL (once I'd written the BCPL compiler).
I managed to subvert the operating system to the point where I could write absolutely anywhere in memory, without privileges (I did read the source code, which I stole). This was used for a variety of nefarious purposes (e.g. remote Ctrl-C equivalent on someone else's terminal, and injection of card images into the batch stream). See also my Anecdotes page.

Honeywell DDP-516 standalone

University of Kent, 1972-1973.
A 16 bit, word oriented machine
Peripherals included a teletype, a printer, and a ½ inch magnetic tape drive, as well as a paper tape reader and punch.
I worked on this for my final year undergraduate project. I rewired parts of the backplane (and added some new boards) to modify and augment its instruction set. This made it more suitable for multiprogramming. For more details, see my Anecdotes page.
All programming in assembly language.

Honeywell DDP-516 under ADMOS

University of Kent, 1972-1973.
A 16 bit, word oriented machine.
Peripherals included a teletype, a printer, and a 0.5 inch tape drive, as well as a paper tape reader and punch. Later, it acquired a disk (see below).
A postgrad named Alan Ball wrote a magnetic tape based operating system called AMOS (Alan's Magtape Operating System). Later we got a disk (2.4MB) and it became ADMOS, for obvious reasons. ADMOS had a very basic filing system, where each user had up to 80 files - named 1 to 80 - resulting in carrying around pieces of paper in one's back pocket, as a reminder of which files were which.
All programming in assembly language. I also ported ML/I to it in my spare time.
My undergraduate final year project involved modifying the CPU to upgrade its instruction set. This involved making new circuit boards and lots of wire wrapping. For more details, see my Anecdotes page.

ICL 1902S under Operator's Executive

Advance Linen Company, 1973-1974.
A 24 bit, word oriented machine.
I was actually a computer operator for two periods of time (vacations). We had small exchangeable disks, but mainly used punched paper tape (some of it being 5 track Telex tape rather than the more common 8 track tape), and magnetic tape. The operator's console was a teletype.

ICL 1904A under Operator's Executive

University of Sussex, 1973-1974.
A 24 bit, word oriented machine.
This had small exchangeable disks, and magnetic tape drives, but mainly used punched cards, some paper tape, and a printer. The operator's console was a teletype.
I did a little bit of work as a computer operator, and some programming in PLAN (the assembly language).
We used an addition to Operator's Executive which was more lightweight than GEORGE (see below). It was called SOSAGE - Sussex Operating System Abandoning George Entirely.

ICL 1904A under GEORGE {2,3}

University of Sussex, 1973-1974.
A 24 bit, word oriented machine.
This had small exchangeable disks, and magnetic tape drives, but mainly used punched cards, some paper tape, and a printer. The operator's console was a teletype.
I did a little bit of work as a computer operator, and some programming in PLAN (the assembler).
GEORGE 2 was a batch management system, and not too bad. The machine was not really big enough for the much more sophisticated GEORGE 3, and they were still running trials when I left.

CTL Modular One under EXEC 2

University of Sussex, 1973-1974.
This just had a teletype, a punched card reader, and a printer. A 16 bit machine.
The machine was the local component in a remote job entry link to a big powerful regional machine in London, a CDC 6600 (which was fairly unusual in that it used ones complement arithmetic). I operated the CTL machine, but had no chance to program it.

Digital DECSystem-10 (PDP-10) under TOPS-10

University of Essex, 1973-1974.
This was a big machine with disks, tapes and DECtapes, as well as an extensive number of teletypes. It also had punched cards and paper tape (the unusual fanfold variety peculiar to Digital and one or two others), and printers. The teletypes were attached via multiplexers which were PDP-8s literally 'bit banging' to drive the asynchronous teletype lines. It had a 36 bit word, an incredibly orthogonal instruction set, and variable size bytes!
It had a 36 bit word, yet retained much of the heritage of the smaller PDP machines. In particular, some models (including this one, a KA-10) had an operator console with a 36-bit switch register (i.e. 36 switches).
I was an M.Sc. student, and used this machine for the year. I mainly used the timesharing part of TOPS-10 (I remember I started with monitor version 506027, and also remember that my user ID was 2600,2645 - funny what the memory retains!)
I programmed this machine in assembler, BCPL, POP-2, LISP, FORTRAN, BASIC and probably some other languages. I also ported ML/I to it in my spare time.

Digital PDP-11/20 under DOS/BATCH (DOS-11)

University of Kent, 1974-1977.
This was a 16 bit byte oriented machine, with 56k bytes of memory (the strange number is because the top 8k of the address space was I/O space). It had two DECtape drives, and a 512k fixed head disk drive, as well as paper tape reader and punch, a teletype, and an early dot matrix printer.
The operating system was DOS/BATCH. This was a basic command line system, occupying a mere 4k bytes of resident memory - the rest was in tiny "overlays" brought in from disk as required.
I programmed this machine in assembly language.

Digital PDP-11/20 under RT-11

University of Kent, 1975-1977.
This was a 16 bit byte oriented machine, with 56k bytes of memory (the strange number is because the top 8k of the address space was I/O space). It had two DECtape drives, and a 512k fixed head disk drive, as well as paper tape reader and punch, a teletype, and an early dot matrix printer.
RT-11 was a "real time" operating system - small, but faster and easier to use than DOS-BATCH.
We didn't have an official copy, but I had the entire source code - I built tools to build it, and make a bootable DECtape to bootstrap it onto the PDP-11/20.
RT-11 ran two 'jobs', so although it was single user, one could run something else in the background (I wrote an automatic print spooler, because the printer was very slow).
I programmed this machine in assembly language.

Digital PDP-11/40 under MUSS-11

University of Kent, 1975-1977
This was a 16 bit byte oriented machine, with 56k bytes of memory (the strange number is because the top 8k of the address space was I/O space). It supported multiple base+limit memory management registers, and had two CPU modes - privileged and non-privileged. It had a disk drive, a teletype, and a spare serial port. It was minimal because it was a development machine.
MUSS-11 was a basic message passing operating system developed at the University of Edinburgh. It used a language called IMP, and had a compiler for it, of course. I was asked to port the IMP compiler to DOS/BATCH. This involved:
- Learning the system, and IMP.
- Writing a file transfer program, using my own error-detecting protocol, to move files to the PDP-11/20 over a serial line.
- Writing shims for the compiler so that it would run on DOS/BATCH.
All this was in preparation for using IMP as a development language for YAROE (see below). In the event, it was not used, but I learned a lot - and I was to encounter IMP again, as well as 'Son of MUSS-11' soon.
I programmed this machine in assembly language and IMP.

Digital PDP-11/40 under UNIX Sixth Edition

University of Kent, 1975-1979
This was a 16 bit byte oriented machine, with 120k bytes of memory (the strange number is because the top 8k of the address space was I/O space). It supported multiple base+limit memory management registers, and had two CPU modes - privileged and non-privileged.
It had a serial line multiplexer (8 lines), a console (a 132 column DECwriter, which was a fast dot matrix printer with a keyboard), and two 2.4MB exchangeable disk drives.
This was only the second UNIX system in England, and the software was delivered on three 2.4MB disks (the manual pages were all on the third disk). I was one of the first half dozen or so users on it.
I programmed this machine in assembly language, and in BCPL (once I'd written the compiler). I also ported ML/I to it in my spare time.

Digital PDP-11/03 under YAROE (Yet Another Rewrite Of Everything)

University of Kent, 1976-1978
This was a PDP-11/03 with just a serial line multiplexer (or two), a paper tape reader and punch, and a connection to our Cambridge Ring network. Its purpose was to allow one to use the same terminal to talk to one of the two ICL 4130 systems, or the new ICL 2960 system. I wrote the user interface (very human friendly!) and the driver software for the paper tape reader and punch.
The system was programmed in BCPL (compiled on a PDP-11/34 and downloaded via a serial line).

Digital PDP-11/34 under DEIMOS

University of Kent, 1978-1986
There were several DEIMOS systems around, mostly performing dedicated tasks; one connected the ICL 2960 mainframe (see below) to the campus ring network, another ran a remote job entry system to the University of London Computer centre, etc.
The system was programmed in the language IMP77, a high level system programming language. A derivative of DEIMOS has been started in C; see http://phobos.tavi.co.uk.
DEIMOS was a microkernel style of system with everything implemented as independent tasks which were able to share memory if desired; this was necessary due to the 16 bit nature of the machine and the resultant per-task memory limitations. It was a derivative of MUSS-11 (see earlier).

Transam Tuscan under CP/M

University of Kent, circa 1979
This was a Zilog Z80 based desktop system with two 51/4" floppy disk drives. CP/M was the most popular 8 bit operating system of the day, and was remarkably functional.
This system was programmed principally in Z80 assembler, although I did implement an editor on it which was written in BCPL and cross-compiled from another system.

Zilog Z80 under Son of YAROE

University of Kent, circa 1980
These systems (and there were several of them) ran essentially the same code as YAROE, apart from hardware specific areas. The BCPL compiler was initially done as a final year project by a student at the University of Cambridge, although I enhanced it a great deal to make the generated code as tight as possible (the Z80 had limited memory).
The systems did the same job as YAROE, although the selected 'mainframe' systems were now various UNIX systems and the ICL 2960 mainframe.

Links

None yet.

ICL 2960 under VME/K

University of Kent, 1976-1979
The ICL 2960 was a low end member of the first series of ICL 2900 ("New Range") machines (this was the "P" series, later succeeded by the rather less complicated "S" series). The low level architecture is actually quite clean and easy to use.
The 2900 series was intended to run an operating system known as VME. There were two variants, VME/B being too big and slow for the 2960, so it used a smaller version known as VME/K. Unfortunately, VME/K was riddled with bugs, and had a mean time between failure (averaged over a 13 week period) of approximately 20 hours!
I did a lot of development and testing on this system, including helping in the development of an intelligent batch scheduler, and writing a BCPL compiler and run time system. I programmed mainly in assembler, with some ALGOL 60, BCPL and FORTRAN.
In 1978, ICL announced that the next version of VME/K would be slimmed down and have fewer facilities; as such, it was no longer fit for purpose as a university system. I spent about a year evaluating EMAS as an alternative (see below), followed by six months planning and implementing the migration to EMAS. This took place in December 1979.

ICL {2960,2970,2972,2976,2980} under EMAS

University of {Kent,Edinburgh}, 1978-1986
The ICL 2960 at Kent was a low end member of the first series of ICL 2900 ("New Range") machines (this was the "P" series, later succeeded by the rather less complicated "S" series). The low level architecture is actually quite clean and easy to use.
I used the other models at Edinburgh occasionally.
Testing on EMAS started in April 1978, and the changeover was made in mid December 1979. I was involved mainly in getting the system set up, users accredited, etc. although I also helped with migration of data from the VME/K disks.
The main problem encountered during the testing phase showed up when a lot of different users logged on and off over a short period of time. This caused a partial flush of a user directory cache, and this triggered a stuck semaphore. This had never been encountered at Edinburgh because of different usage patterns and a smaller total number of registered users. I actually fixed the bug before Edinburgh worked out the problem!
The mean time between failure (averaged over a 13 week period) was about 2000 hours - compare this with the VME/K figure noted above!
I did a lot of development on this system, mainly in Edinburgh IMP (it was written in that language); I also used BCPL, as I had implemented a compiler for it. The development included work on practically every component of the operating system, as well as numerous applications.
At one point we had a persistent problem with the machine microcode halting. I taught myself the microcode, and read the source code to find the problem. This turned out to be a design flaw in the hardware (never acknowledged by ICL). I fixed this by writing a microcode patch (hand assembled) to generate a new type of exception, and then writing tools to insert it into the system. I then implemented code at all levels in the operating system to pass the new exception code back to user level. See my Anecdotes page for more details.
In 1983, we acquired a 'free' second Order Code Processor or OCP (more commonly known as a CPU). The operating system was capable of proper symmetric multiprocessing (SMP) but the precise hardware control interface specification was unavailable (presumably ICL was either secretive or incompetent - I favour the latter explanation). The location and layout of the appropriate 'image store' control registers was unknown. I went back to the microcode source, and after three days of reverse engineering I had generated a single sheet of A4 containing the necessary specification. I then modified the operating system using this, and we had full SMP! See my Anecdotes page for more details.

ICL 2970 under DME and GEORGE 3

British Telecom, 1984-1985
I was commissioned to implement ML/I on this system. It was used to generate test data for System X.

ICL {2956,2980} under VME/B

Dun and Bradstreet, 1987-1989
I was commissioned to write a native compiler for PDL, which was essentially a database query language. This was already running on the system, but was not fast enough (it was interpreted). In order to sell it to a very large customer (hardware costs of £40 million were mentioned), it had to be speeded up. I achieved this by writing a compiler which generated native machine code; the original used an interpreted pseudo-code. This was all quite challenging because both the original compiler, and the interpreter, were written in COBOL. The interpreter contained a lot of COBOL support code which had to be incorporated into a new run-time system (written in assembly language).

Links

None yet.

Digital VAX {750,780} under 4.{0,1,2,3}BSD

University of Kent, 1978-circa 1986

Links

None yet.

Digital VAX {3100,8200,8800} under VAX/VMS {3.x,4.x,5.x }

University of Kent, 1983-1994

Links

None yet.

Digital VAX {2000,3100,4000} under VAX/VMS {7.3}

Personal, 2008-present

Links

None yet.

BBC microcomputer

University of Kent, 1983-1985

Links

None yet.

Olivetti L1 under some kind of UNIX

Olivetti, Milan, 1985

Links

None yet.

AT&T 3B2 under UNIX System V

Olivetti, Milan, 1985

Links

None yet.

IBM 370 under MVS

RHM Foods, Harlow, 1986

Links

None yet.

IBM 4381 under VM/XA, CMS and EMAS

IBM Greenford, 1986

Links

None yet.

Digital PDP-8 clone standalone

Personal, 2009-present

Digital PDP-8 clone under OS/8

Personal, 2009-present

Advance 86B under MS-DOS 2.0

Personal, 1984-1985

IBM 5150 (original PC) under PC-DOS 2.0

IBM South Bank, 1984

Comcen PC under MS-DOS 3.0

Personal, 1985-1989
I couldn't find a picture, but it looked pretty well exactly like the 5150, although it had a clone motherboard that was just like the one in the PC/XT.

Links

None yet.

IBM PC Portable under PC-DOS 3.0

Personal, 1986-1992
I still have one of these somewhere.