Timeline of the MUSE/Atlas project at Manchester University, 1955 - 1971
Simon Lavington
5 December 2012
Acknowledgements
The reminiscences of many former Atlas pioneers have contributed to the compiling of
these notes. Particular thanks are due to Dave Aspinall, Tony Brooker, Yao Chen, Eric
Dunstan, Dai Edwards, Iain MacCallum and Robin Kerr.
- 1954, spring:
- The new Dover Street building is occupied by the University's
Department of Electrical Engineering. Soon afterwards, the Ferranti Mark I is moved from
the Coupland Street building to Computing Machine Room 1 on the third floor of the Dover
Street Building. Meg (the prototype Mercury) is moved to Computing Machine Room 2.
After re-commissioning and elimination of interference from radar being tested to
destruction from a nearby Manchester location, local users were running programs on Meg
by early 1955. Meg was also used by a visitor from the Norwegian Atomic Energy
Organisation for about 6 months. The Norwegians then purchased the first production
Ferranti Mercury, which was based on Meg.
- 1955, autumn:
- The computer research group within Electrical Engineering, led by
Tom Kilburn, started to investigate the possibilities of higher speed transistor logic circuits
and faster memory systems of improved capacity. This work was intended to focus
thoughts for the development of a new high-speed computer but there was no specific
intent to build or even to specify a machine at this time. There was, however, the thought
that transfers between the primary memory (probably ferrite cores) and backing store
(probably a magnetic drum) could be automated by hardware so as to relieve users of this
task. Tony Brooker had achieved some software support for this process on the Ferranti
Mark I computer, within the Mark I Autocode system. No firm strategy for hardware
memory management had yet been devised for a new machine.
- 1956 Feb:
- Ferranti's computer department moves into new premises at West
Gorton, Manchester.
- 1956 Dec:
- The design objectives for the Univac LARC and the IBM STRETCH
computers are presented at the Eastern Joint Computer Conference in the USA.
- 1957 Jan:
- Jack Howlett (AERE Harwell) and J Corner (AWRE) produce for the
UKAEA a report entitled The case for a big computer, stating that 'by 1960 the AEA would
need a computer comparable to LARC and STRETCH'. In response to the Howlett/Corner
report, the National Research Development Corporation (NRDC) Computer Sub
Committee asks Lord Halsbury to review and compare US and UK scientific computer
development activity. From about this time, the word 'supercomputer' starts to be used by
NRDC. See references [1 to 4] for the background to the NRDC's interactions with the
MUSE/Atlas project at Manchester. The references are listed at the end of this document.
- 1957 early:
- A decision was made within the research group led by Tom Kilburn to design
and build a fast computer with a target speed of a microsecond per instruction. The
project was called MUSE, short for Musecond Engine. This would be a completely parallel
machine and would have to use the latest high-speed germanium junction transistors then
just becoming available. Dai Edwards would be responsible for the engineering aspects
and Tony Brooker for the software. To support the project, a sum of £50K was made
available from the Department's own research fund. This fund had been accumulating
since 1951, as an agreed percentage of the University's overall Computer Earnings Fund.
- 1957 March:
- Dick Grimsdale begins work on a fast fixed store (ROM) system and
Yao Chen on fast RAM storage. For the latter, the possibilities of smaller core sizes and
using two cores per bit and/or partial flux switching were examined, to see what speed
improvements were possible.
- 1957 30th April:
- John Cockcroft calls a meeting of UK computer users and designers
at UKAEA to discuss technical problems and requirements for a high-speed computer.
The meeting was attended by seven people from AERE, representatives from several
other government establishments, and Kilburn from Manchester, Wilkes from Cambridge
and Strachey from NRDC. The meeting made reference to UKAEA's need for 105 mesh-
operations per second on a 3-d mesh for reactor calculations - (a paper entitled The
Three-dimensional Computer had been distributed by AERE). Amongst advanced
technologies discussed were thin magnetic films and diode-capacitance stores, and the
need to develop much faster transistor circuits and much faster input/output. At the
meeting, UKAEA found itself in disagreement with NRDC's proposal that NRDC itself (in
practice, meaning Christopher Strachey) would design the machine.
- 1957 26th June:
- NRDC agrees to put 'not less than £1 million' into a high-speed
computer project. At this point, NRDC felt that the development would be beyond the
capabilities of any one single UK manufacturer. Accordingly, NRDC proposed a joint
project which would (a) have contributions from all the UK's centres of expertise, and (b)
would give NRDC an opportunity to gain some return on its investments. In the event, it
proved impossible for NRDC to satisfy (a) and (b) together. (Note that by this time, IBM
was said to be spending $28m annually on STRETCH and to be deploying 300 graduates
on the project).
- 1957, late autumn:
- Ferranti Mercury computer is installed in Computing Machine Room 2
on the third floor of the Dover Street building. Mercury had passed its acceptance tests by
5th February 1958.
- 1957, late:
- Dave Aspinall, seeking a project for his Ph.D., decides not to work on
thin magnetic films for Pete Hoffmann but to join the MUSE team and start work on the fast
parallel adder. Test circuits were built using the latest American germanium 2N501 and
SB240 high-speed transistors - each costing about £35 and therefore much too expensive
for all but the time-critical carry-path. With its better symmetry and lower saturation
voltage, the SB240 proved more satisfactory for the carry-path. Cost considerations and
supplier-availability caused Kilburn's group to select Mullard OC170 transistors for the
main MUSE logic. These graded base transistors were fast if operated with a significant
voltage between emitter and collector at all times but could not be allowed to saturate
without significant loss of speed. A non-saturating, defined-current, circuit was devised for
MUSE, with clamping diodes limiting the output voltage levels. The cable driver and cable
receiver circuits used on MUSE for inter-unit connection were also interesting. 8 milliamps
was pumped through the standard OC170 circuit down a piece of miniature co-axial cable
and regenerated into a standard signal at the other end.
- 1958 27th Feb:
- Meeting (later known as the Second Harwell Computer Conference)
between NRDC and representatives from AERE, RRE, NPL, MOS and the Universities of
Cambridge and Manchester. This meeting was initiated by Sir Owen Wansborough-Jones,
then Chief Scientific Adviser to the government. A specification based on general
requirements for a high-speed computer was circulated by Cooke-Yarborough (AERE).
The meeting recommended the development of two computers: (a) a shorter-term project,
having a fast store of 105 numbers and 105 mesh-operations per second, to be available
commercially in five years' time; (b) a longer-term project involving new components and
new techniques, for which the design would not start for another 3 to 5 years. The
meeting recommended that a Working Party consisting of Kilburn, Wheeler, Strachey,
Howlett and Cooke-Yarborough be set up to check the suitability of MUSE for (a), the
short-term project, and to report in a months' time. Kilburn contributed to the Working
Party's discussions but, for obvious reasons, did not participate in the writing of the final
Report. On a visit to Manchester, which took place on 17th March, the Working Party were
shown the MUSE developments, which by then included about 500 words of the MUSE
fixed store with an access-time of 0.2 microseconds, one word of a two-core-per bit RAM
with a cycle time of 2.5 microseconds, and two single-bit experimental adders using
different types of transistor. (There was an extreme shortage of suitable transistors at that
time and they were very expensive).
-
The Working Party noted that the Manchester project "aimed at producing within three or
four years a computer with times for basic arithmetical operations of a few microseconds
using components already in existence. It should be pointed out that the information
quoted here has been disclosed at a much earlier stage than Dr Kilburn would have
wished had he not been specifically asked".
-
Amongst the conclusions in the Working Party's final Report were the following points:
-
"Dr Kilburn has in existence at Manchester a very lively team of experienced
computer circuit engineers. They have close and easy relations with Ferranti
Ltd. - [However] we doubt whether even this combination has the strength and
breadth needed for execution of the stage 1 project unaided".
-
"The speeds of the proposed machine [MUSE] fall somewhat below the target
originally suggested [of 105 mesh operations per second], but this is not a very
serious matter".
-
"The proposed machine [MUSE] falls short of the minimum requirements for the
stage (a) project in respect of storage capacity facilities and provision for
peripheral equipment. The concept as it stands might be scaled up, but this
would not provide a machine of the type we [particularly the UKAEA] require and
one which is suitable for commercial manufacture. A fresh logical design would
be necessary".
-
"Dr Kilburn made it quite clear that the present University project was not
dependent upon support from NRDC and would be completed without it".
-
"Dr Kilburn further made it clear that he would only accept [responsibility for
building] the [stage (a)] project on the condition that he was to be the final arbiter
on all questions of engineering and logical design, although he would be willing
to listen to suggestions from other people during the first 18 months".
-
"Taking the above points into consideration - we cannot recommend that the
Manchester University Project should be the sole basis for the Stage (a)
computer".
-
"We therefore recommend that the full meeting [ie between NRDC and
representatives from AERE, RRE, NPL, MOS and the Universities of Cambridge
and Manchester] be reconvened to consider whether it would be possible to
combine the Manchester work with that of other organisations [for example
RRE] or individuals or, failing this, what other course should be adopted".
1958 Oct. After various meetings and conflicting advice, NRDC renewed its
approaches to EMI, English Electric and Ferranti in an attempt to inject funds into a
company that would undertake a British supercomputer project. English Electric
responded negatively, and EMI positively, over the terms of NRDC's loan. An EMI pilot
project was eventually given NRDC support but it came to nothing. Ferranti did not agree
with the terms of NRDC's offer of a loan.
- 1958 Oct:
- During the summer period, Tom Kilburn had had private discussions
with Sir Vincent de Ferranti. It is deduced that these talks were fruitful. In October Ferranti
decided to go ahead independently of NRDC with a cooperative project with Manchester
University that thereupon changed its name from MUSE to Atlas. This decision was
confirmed in January 1959. (The slight uncertainty about informal versus formal
agreement may have arisen because Brian Pollard resigned as Manager of Ferranti's
Computer Department and Peter Hall took over in December 1958).
- 1958 24th Dec:
- Ferranti Mark I shut down. It was dismantled in June 1959 and the
bits given to Ferranti Ltd. Once cleared, Computing Machine Room 1 was assigned to
house the so-called Pilot Model of the MUSE/Atlas project.
- 1959 Jan:
- Issue 2 of the document Provisional users' code for the Manchester
University computer, MUSE is circulated.
- 1959 Feb:
- (a). The first paper describing the Atlas parallel adder is presented at
the Institution of Electrical Engineers, London [ref. 5]. (b). Strachey files a patent for 'timesharing
in large computers' (British Patent 924672, granted in 1963). In 1959, 'timesharing'
was the term used in the UK for what we would now call 'multiprogramming'.
Strachey's patent appears to make 57 claims, specifying a director (supervisor) in fixed
store, a memory-protection mechanism and an interrupt priority structure. Note that
Kilburn's group at Manchester University filed 15 Atlas patents between October 1957 and
February 1962 through NRDC, on subjects including arithmetic units, magnetic recording,
storage systems and memory management -(see also below under March 1960).
- 1959 May:
- NRDC loaned Ferranti £300K towards the cost of Atlas
development, to be recovered from sales. Ferranti had paid off their loan by March 1963,
according to [ref. 4].
- 1959 15th/20th June:
- IFIP UNESCO Computer Conference in Paris: separate
papers on high-performance computers are presented by Kilburn and by Strachey - see
[ref. 6 and 7]. The appearance of Strachey's paper was a complete surprise to the
Manchester team who, by contractual agreement, had always discussed their computer
research as early as possible with NRDC to ensure that it was covered by patents. With
the benefit of hindsight, it seems that some of the ideas in Strachey's paper could be said
to have been inspired by - or shared by - the MUSE/Atlas project. These include: the
precise speed and use of a Fixed Store; the use of private RAM for the 'Director'
[Supervisor] working area; the arithmetic instruction speeds; the use of fixed-block
magnetic tapes; the need for some kind of job control language. Two of Strachey's ideas
that were definitely not in MUSE/Atlas include: the provision of two extra consoles (one for
maintenance engineers, the other for program testing) - these were in addition to the main
operator's console; the use of datum/limit registers for memory management.
-
A great many papers on the hardware and software of the Atlas system then appeared in
the period 1960 to 1967. The main ones are listed as [refs 8 to 21] at the end of this
document. It is believed that Christopher Strachey published no further papers on
computer architecture. For an account of his other work, see [ref. 22].
- 1959 19th June:
- Date on which Ferranti starts to issue copies of the massive
manual known locally as the Atlas Bible. This was a foolscap spring binder for containing
confidential technical information, arranged in 13 sections. Ben Cooper of Ferranti was in
charge of organising the circulation and periodic updating. The preface describes the
manual as "a description of the [Atlas] system from the viewpoint of a knowledgeable
programmer, and is intended for the use of persons concerned with the design of the
system". Updates to various pages were issued periodically until at least 13th July 1963.
By this time the manual was over an inch thick.
- 1959 autumn:
- Towards the end of this year, Dexion frames had been
assembled in the old Ferranti Mark I room on the top floor of Dover Street and all the Pilot
Model developments transferred to this room. During 1959 the MUSE/Atlas team started
to lay out printed circuit boards, to be manufactured by Ferranti at West Gorton ready to go
into the Pilot Model racks and into the final production ATLAS logic racks.
- 1959, 9th Nov:
- On this date, formal Atlas hardware progress meetings started
at Ferranti's West Gorton site. The meetings were chaired by Peter Hall. Keith Lonsdale,
Gordon Scarrott and others attended on behalf of Ferranti. Tom Kilburn and Dai Edwards
attended on behalf of the University. These meetings continued thereafter to be held on
the second Monday in the month. At the first meeting it was formally decided that MUSE
should be the Pilot Model for Atlas and that any further MUSE design should as far as
possible be part of the final design for the production Atlas. It was agreed that some of the
drum, magnetic tape and fixed store equipment would be attached to the Pilot Model but
that in general the Atlas team would not be expected to work on peripheral equipment.
Keith Lonsdale said that the design of peripheral equipment would be carried out by
Ferranti's Orion team and the designs made available to the Atlas team. Subsequently
there proved to be significant exceptions to Lonsdale's statement, particularly in relation to
the magnetic tape and magnetic drum equipment. For example, higher drum transfer
rates were needed for Atlas so the Orion revolution time was doubled for Atlas and new
forms of read/write head were devised (see also the note below for May 1960 and
Appendix 2).
- 1959, end:
- On the software side, Tony Brooker, Derrick Morris, Robin Kerr
and Jeff Rohl were planning the writing of high-level language compilers for Atlas. For this
work, Brooker conceived the Compiler Compiler [ref. 18].
- 1960, January:
- The decision was made to sub-contract work on the B store
and RAM main store design and manufacture to Plessey Ltd. This entailed a significant
amount of work by Plessey. It was considered that the B store would probably use two
cores per bit and be partially flux-switched for speed, and that the main RAM would
probably use one core per bit and be constructed in units (stacks) of 4K words of 50 bits
each (48 data plus two parity).
- 1960, 30th Jan:
- The complete 24-bit B unit parallel adder was the first major
section of MUSE/Atlas Pilot Model to be successfully tested on this day. The paper
describing this [ref. 12] was received by the IEE on 28th January and published in
November 1960.
- 1960, Spring:
- Various meeting at which Tom Kilburn and Peter Hall (Ferranti)
assigned formal responsibilities for progressing particular sub-sections of Atlas. Rather
than list names/tasks here, and because the design and development team expanded over
the next couple of years, a full list of names, dates and responsibilities is given at the end,
in Appendix 1.
- 1960 March:
- Three Atlas hardware papers (storage and ALU technologies)
were presented by the Manchester University team at the IEE meeting in London on 1st
March 1960 - see [refs. 9, 10, 12]. The papers subsequently appeared in the November
1960 issue of the IEE Proceedings. Of the three Atlas patents that laid the foundation for
what became known as Virtual Memory, one was filed on 16th March 1960 ('address
translation'), and two on 20th April 1960 ('block transfers' and 'replacement algorithms').
- 1960 May:
- Problems with the mechanical design of the (Orion) drums
became evident. The Atlas team at the University became involved in the design of the
read/write electronics, since Gordon Scarrott (of the Ferranti Orion team) said that there
was no spare effort for this at West Gorton. Also at this time the Lancashire Dynamo
company was commissioned to build the power supply and switching system for the
production Atlas.
- 1960 August:
- Confirmation that designers at Ferranti's Edinburgh factory
were to organise and supply the magnetic tape decks. They chose the Ampex company as
a likely source. The first Ampex deck (type FR300) that was received proved
unacceptable because the acceleration time to operating speed was too slow by a factor of
three. A new Ampex deck, the TM2, was ordered but the quoted delivery-time was one
year. Only the basic Ampex TM2 decks were to be purchased: Ferranti Edinburgh were to
design and supply the read/write electronics and fit them to the decks for both the Orion
and Atlas projects.
- 1960 30th Sept:
- By this date, the outline design of all nine Atlas sub-units had
been completed and the detailed logical design of 80% of these units had been finished.
At this stage there are 49 types of printed-circuit board (pcb) in production or ready for
production at Ferranti's West Gorton factory; there are 9 types for which the design is still
incomplete. The package-count for 60% of the computer had been completed and the
backwiring schedules for 20% of the computer had been produced - see [ref. 23]. The
nine sub-units that are distinguished in [ref. 23] are:
1 CPU: B arith and B store
2 CPU: A arith
3 CPU: Distributor
4 Core store control
5 Drum control
6 Peripheral Coordinator
7 Tape coordinator
8 Fixed store
9 Working store control plus the bought-in store unit.
By 30th Sept. the Pilot Model has the following sub-units in working order: (1), part of (3), a
pilot version of (8), loaded with 850 words of information. A Creed Paper Tape Punch was
connected to the Pilot Model at this stage. The Plessey 1K word store was delivered in
August but not yet tested. Simple test programs have been run on the Pilot Model.
- 1960, 10th Oct:
- At the monthly progress meeting it was stated that successful
operation of the B-store, the 1K working store and the fixed store had occurred.
- 1961, end-January:
- The Dexion Pilot Model of Atlas was dismantled and reassembled
at West Gorton, where it was used for further development and, in due course,
the training of maintenance engineers. At the University, work began on refurbishing the
Computing Machine Room 1 at Dover Street with a false floor, air conditioning ducting, etc.
- 1961 February:
- Ferranti Edinburgh stated that they are short of people to
design the read/write electronics for the magnetic tape decks, though they had the
capacity to manufacture and test. They asked for help, in order to complete the design
and thus to deliver the deck systems on schedule. Two of the Atlas team, Dai Edwards
and David Aspinall, agreed to help. They promptly went to Edinburgh and completed the
necessary design work in one week. Sam Alexander (Ferranti) accepted the design work
and confirmed that the resulting systems would also be suitable for Orion.
- 1961, March:
- Work was advancing at West Gorton on the construction of the
logic cabinets for the first production Atlas - destined for the University of Manchester. At
Dover Street the refurbishment of Computing Machine Room 1 continued.
- 1961, early summer:
- Detailed design of the Atlas Tape Coordinator handed over by
Dave Aspinall to West Gorton for production. Dave Aspinall then took time off to write up
his Ph.D. (awarded December 1961). By about this time, Kilburn's informal estimates
[ref.24] of the constituent parts of the final Atlas appear to be as follows:
Unit Packages Boxes Racks
Central computer (ie A, B&C, D) 1200 30 3
Tape coordinator 350 9 2
Drum cabinet 200 ? 2
Drum coordinator 120 3 1
Core store coordinator 385 8 1
Core store 50 x 4
Working store 50 1 1
Periph coordinator 266 7 1
Fixed store 334 non-std 2
- 1961, June:
- The first Atlas production cabinet, containing the A, B&C and D
units, is delivered to Dover Street and installed in the refurbished Computing Machine
Room 1.
- 1961, 30th Sept:
- By this time, the total number of pcbs in Atlas had increased to
4,403, of which 1200 had been completed and tested, 1,000 awaited testing and the rest
were still in manufacture [ref.23]. More completed cabinets had been delivered to Dover
Street. The final board-count (including I/O peripherals and special online devices such as
the X-ray Diffractometer and the Speech Converter) had risen to 5,257 by 1965 [ref. 25].
This reference also gives a rough estimate of the total number of transistors (60,000) and
diodes (300,000).
- 1961 November:
- By now the manufacture of the NIRNS/Chilton Atlas had begun
at West Gorton [ref. 26]. This probably implies that, by November, most cabinets of the
Manchester Atlas had been delivered to the University. The power supplies were delivered
to Dover Street by Lancashire Dynamo.
- 1962, 29th January:
- The cabinet containing the Tape Coordinator was delivered
from West Gorton to Dover Street.
- 1962 February:
- On 16th February 1962 Kilburn reported [ref. 26] that: "some
components of the Manchester Atlas were now being tested, and it was already clear that
the speed of the arithmetical processor was equal to expectations".
- 1962, March:
- The detailed logical design of Atlas was declared to be
'complete' by this time [ref. 23]. 4,065 pcbs had been delivered to the University. The first
half of the fixed store was working and had been loaded with information. The delivery of
the first drum unit was expected by early April. Iain MacCallum, then an M.Sc. student
working on the Compiler Compiler, remembers "the great excitement when the first
'extracodes' were tested on the fixed store memory. We wondered: was this the fastest
executed programs ever?" [ref. 27].
- 1962 early June:
- The Compiler Compiler becomes available. As far as Atlas
software is concerned, the first language to which the Compiler Compiler is fully applied is
Extended Mercury Autocode, EMA. By about the end of July EMA became available to
end-users such as Kodak (see below). This helped users in the transition from Mercury to
Atlas.
- 1962 July:
- On 6th July Kilburn reported [ref. 26] that: "the central computer
continued in use for the development of programmes for the Supervisor. Operating times
were proving to be remarkably close to those predicted. Both halves of the core store had
passed their acceptance tests at Plessey's. The core store, the drum store and one tape
unit would be linked into the central computer via their respective co-ordinators during the
coming fortnight, and the Atlas should be in full operation on internal programming work
during August".
- 1962 summer:
- By the summer of 1962 there were several of Ferranti's
Computing Service customers who were anxious to begin running their applications
programs on Atlas. Companies with urgent projects, such as Kodak and Royal Dutch
Shell, were invited to use the Manchester Atlas on an 'at risk' basis whilst the Supervisor
was still being developed. Furthermore, these 'apps programmers' were asked if they
would like to help iron out bugs in the Supervisor. Ann Moffatt from Kodak, who used the
Manchester Atlas from about July to December 1962, remembers that: "We were given
fairly detailed talks on how the Supervisor was supposed to work. I was given the 'drum
learning program' to test and debug. Other customer programmers took other bits to test
and debug. The programmers who had been sent by their companies to help get Atlas
working were the best their companies could find. We had a healthy respect for each other
and worked well together, excited to be working on the most powerful computer in the
world at that time- We could use the machine between 11:00pm and 6:00am. I was
writing my application programs in Mercury Autocode."
- 1962 Sept:
- Internal course on programming in Atlas Autocode held for staff
and research students in Lecture Theatre A, Dover Street.
- 1962, autumn:
- Commissioning of Atlas progresses at Dover Street. David
Aspinall remembers that: "At one stage, when Ferranti were having problems with the
drum, one tape deck was programmed to emulate the drum store. Eventually, by
December 1962, all eight magnetic tape channels were commissioned". Also, it was
reported [ref. 26] on 27th November that some work had been done on an experimental
data link between Manchester and Edinburgh. At this stage the links were off-line,
requiring paper tape input/output at both ends and good human communication to ensure
satisfactory operation. On-line data links did not become available until about 1965. Also
at the same meeting, Anelex lineprinters were described as newly-available - (see also
under March 1963).
- 1962 Oct:
- Iain MacCallum and John Clegg joined Robin Kerr to start work
on the Atlas Algol compiler.
- 1962, 7th December:
- Sir John Cockcroft officially inaugurates the Ferranti Atlas in the
Dover Street building, accompanied by press coverage and demonstrations of the
computer.
- 1963 Jan:
- The Mercury computer is moved out of Dover Street and
thence to Sheffield University, by prior arrangement. The Mercury room, Computing
Machine Room 2, is then used for Atlas data preparation equipment and for the online X-
ray Diffractometer - (see below). From January 1963 the University of Manchester's
Computing Service is run on Atlas. See below (June 1963) for the final form of the
Agreement which gave the University half the available time on Atlas.
- 1963 February:
- First bound copies of the Atlas Autocode Programming Manual
issued. Authors: R A Brooker and J S Rohl.
- 1963, March:
- An Anelex 1,000 lines/minute lineprinter, a Creed 3000 paper
tape punch (300 chars/sec.) and a TR7 paper tape reader (1,000 chars/sec.) were added
to Atlas to complement the existing slower I/O peripherals. The lineprinter replaced an
experimental graphical output device. At some time before November 1966 a second
Anelex lineprinter had been added.
- 1963, spring:
- On 29th May meeting at NIRNS [ref. 2] Kilburn reported that:
"with the qualifications concerning the drum store which had already been discussed, the
Manchester University Atlas was complete and maintenance was now in the hands of the
Ferranti engineers. The Mercury Autocode [compiler] was finished; the Atlas Autocode
[compiler] was in use but he would expect some faults to be found in it for a few weeks".
- June 1963:
- By letter dated 3rd June 1963 to R A Rainsford (the University
of Manchester's Bursar) from Peter Hall (Manager of Ferranti's Computer Department), the
final agreement regarding the ownership and operation of the first production Atlas at
Manchester University is as follows:
- The Atlas computer is owned by Ferranti Ltd.
- This Agreement is for 8 years, starting 1st Jan 1963.
- The University guarantees Ferranti unrestricted access to the machine.
- Ferranti will maintain the machine for at least the equivalent of one shift without
charge, and intends to maintain the machine on a three-shift basis.
- Ferranti will make available to Manchester University (MU), without charge, half the
capacity of the computer after deduction of (i) time wasted due to system failures;
(ii) engineering maintenance time.
- Ferranti will provide the operating team.
- Ferranti will pay MU £10K per annum for use of building, power, etc.
- Ferranti will pay MU £100K in consideration of MU's technical assistance with the
Atlas project. Additionally, Ferranti will pay MU 5% of the invoice value of Ferranti
Alas Computing Service external jobs executed during the period of the agreement.
This Agreement was approved by the University Council in July 1963.
- 1963 26th Sept:
- ICT acquires Ferranti's mainframe computer interests. Shortly
afterwards, Atlas hardware and literature start to carry the ICT logo.
- 1963, Oct:
- Tony Brooker returns from a year spent at IBM's Yorktown Heights
Research Laboratory, New York. He wrote a conventional compiler for Atlas Autocode in
the Autumn term of '63 as a back up for Jeff Rohl's Compiler Compiler version "in case
for whatever reason it proved too slow. In fact both versions were used but I was able to
twig mine to provide statistics about the relative usage of different statements". [ref. 28].
- 1963, 11th December:
- Bart Fossey and Bob Hopgood from Harwell visit the two Algol
teams at Manchester - the main (ICT) one led by Robin Kerr/John Clegg and a second
(University) project led by Derrick Morris.
- 1963, end-December:
- The first version of the Algol compiler was in use and the
design of the second version was complete. This second design required a new feature in
the Compiler Compiler, which Derrick Morris completed early in 1964. Iain MacCallum left
the Algol team at the end of 1963 [ref. 27].
- 1964, January:
- The full Supervisor becomes available.
- 1964, spring:
- The on-line X-ray Diffractometer becomes operational. The
enhanced version of this, which was implemented for a PDP8 computer, is described in
[ref. 29].
- 1964, May:
- The second version of the Algol compiler is completed.
- 1964, spring:
- On 5th May it was reported [ref. 26] that tests on the
NIRNS/Chilton Atlas at West Gorton "indicated that computing times were within those
specified with the exception of the time for division which ranged from 10 -30 microsecs
with an average of 25.6 compared with a specified average of 17.9 microsecs. Professor
Kilburn said that this was due to a recognised logical error in the divider which he had
decided not to rectify because the time taken to do so would have been considerable
whereas the effect on the overall speed of the machine was small. The overall speed
originally expected had been 60 -80 times that of Mercury and the overall speed now
found was 70 -80 times that of Mercury".
- July 1964
- Robin Kerr departs for the USA, leaving John Clegg in charge of Algol
developments. During 1964 the Harwell Algol group collaborated with John Clegg's team.
- 1964
- Tom Kilburn's group splits off from Electrical Engineering to form a
separate Department of Computer Science, initially still located in Dover Street. The initial
complement is 12 academic staff.
1964, December The Speech Converter comes into operation. This is an on-line 8-bit
analogue/digital/analogue converter with 20KHz bandwidth, used primarily for research
into speech analysis/synthesis and into automatic speech recognition see [ref. 30]. It is
also used for other signal analysis applications.
- 1964, end:
- By now the Computing Service within the Department of Computer
Science was staffed by one 'lecturer-grade' person plus about 14 people of
technician/secretarial grade. About 40 to 50 University Departments or Departmental-
groups were catered for, plus about 15 outside (academic) users. It is estimated that the
computing needs of the University of Manchester, including UMIST, will exceed the time
available on Atlas by 1967 [ref. 31]. Hence a proposal is put forward for a new building to
house both the Department of Computer Science and the University's Computing Service,
together with provision of a new 'Service' computer costing about £2m.
- 1965, early:
- The on-line data link becomes operational - [ref. 32].
- 1965, 1st March:
- Date of issue of an Atlas Autocode Reference Manual (authors
listed as R A Brooker and J S Rohl).
- 1965, June:
- From time to time the Computing Service re-adjusts the
charging formula used to assign costs (and hence resource-allocations) for outside users
running programs on Atlas. On 18th June the charging is calculated as follows. The total
cost (in pence) of a job is given by the following formula, which has four components:
Computing = A/10(1+n/100)
Input = 36B + 25.5C
Output = 0.5D + 2E + 60F
Mag tape = J/40 + nT/10 + m(240 + 2T + A/100)
Where:
A = instruction Count
B = PTR input (blocks)
C = CDR input (blocks)
D = LPT output (records)
E = CDP output (records)
F = PTP output (blocks)
n = requested store (blocks)
J = mag tape transfers (blocks)
T = mag tape half time (secs)
m = number of mag tape decks used
G = number of drum transfers.
Also, the computing time in secs = 0.006144A OR 0.011G (whichever is the greater), + 90 if n>100.
- Note: the above cost is full rate; academics (from other institutions) are charged 20% of
this rate.
- 1965, autumn:
- The Department of Computer Science moves into temporary
accommodation in Coupland Street, leaving the Computing Service in Dover Street.
- 1965 October:
- The Department of Computer Science accepts its first intake of
undergraduates into a three-year honours degree course.
- 1966:
- Tom Kilburn's research group within the Department of
Computer Science starts plans for a new high-speed computer, targeted at being about
twenty times faster than Atlas. The machine was later called MU5.
- 1967, 1st October:
- A 16M word Data Products type 5048 disk store with two
independent read-write mechanisms, recently added to Atlas, comes into service on this
day as a file store. The disk in situated in Computing Machine Room 2.
- 1968, July:
- A Computing Service analysis of the use by academics of Atlas
during the period July 1967 to July 1968 [ref. 31] finds that 109,200 programs have been
run, with the following statistics:
User group |
no of Depts |
Equiv. cost of computing (£) |
Total time (hrs) |
no. of jobs |
Manchester Univ. |
43 |
478,700 |
2880 |
69,900 |
UMIST |
18 |
251,100 |
1640 |
37,600 |
Other Univs |
8 |
4,400 |
20 |
500 |
Research Orgs |
6 |
14,200 |
110 |
1,200 |
Totals |
75 |
748,400 |
4650 |
109,200 |
(Of course, the figures for the Ferranti Computing Service and commercial customers are
excluded from the above analysis).
- 1968 Oct:
- ICT merged with English Electric Computers to become ICL
(International Computers Limited).
- 1968:
- The Science Research Council (SRC) awards Tom Kilburn's research
group a five-year grant of £630,446 for the MU5 project [ref. 33]. ICL
West Gorton agrees to give significant help with production facilities. A
further SRC grant of £130,000 for MU5 was received in 1973.
1969 autumn The Department's Computing Service section is taken over by the
government-funded University of Manchester Regional Computing
Centre (UMRCC). Gordon Black is appointed as the first UMRCC
Director. The Centre is temporarily located at UMIST and an ICL
1905F is chosen as an interim facility. Meanwhile, a new Computer
Building is designed to house both the academic and the service
sections.
- Spring 1970:
- UMRCC starts running an interim computing service, based on an ICL
1905F computer installed in the Maths Department at UMIST.
1971: By now, UMRCC had selected a CDC 7600 computer front-ended by
an ICL 1906A as their main service facility. The two machines are
first installed and tested at West Gorton, since the University's new
Computer Building is not yet complete. By now, the MU5 design team
numbered 16 Computer Science staff backed by about 25 research
students and 19 engineers seconded from ICL West Gorton.
- 1971, 30th Sept:
- The Manchester Atlas is switched off. The console, an Ampex
tape deck, a small section of the Fixed Store and the logic bay
containing the Page Address Registers are preserved for display in
the Department of Computer Science. Miscellaneous documentation,
including a copy of the Atlas Bible and a mag tape containing the
Supervisor, is preserved. The remaining equipment is sold for scrap.
Postscript
- 1972 autumn
- A new Computer Building, subsequently to be known as the Kilburn
Building, is opened. UMRCC occupies the ground floor and the
Department of Computer Science occupies the rest. UMRCC installs a
CDC7600 computer front-ended by an ICL 1906A. UMRCC initially
provided services for a group of universities which included Salford,
Liverpool, Keele and Lancaster. UMRCC equipment was then
updated over the years. For example a Cyber 205 vector
supercomputer was commissioned in the mid-1980s, to be followed in
due course by a Fujitsu VP1200 and VPX 24/10. The number of
universities served expanded so that it became a national service.
The name changed from UMRCC to Manchester Computer Centre
(MCC) and finally to just Manchester Computing (MC) when it was
wound down in the mid-2000s, by which time each UK university was
making its own provisions. The University of Manchester's current
Directorate of IT Services was established in 2005 just after the
merger with UMIST. At the time of writing, the Directorate still
operates from the ground floor of the Kilburn Building and with IT
teams right across the campus.
- 1974 October
- ICL announces a new range of computers called the 2900 Series.
The architecture of the 2900 Series owes much to and has a great
deal in common with MU5. By October 1974 MU5 was running Algol
and Fortran benchmarks and the assessment of hardware and
software performance had begun.
- By now the Department of Computer Science numbers 32 full-time
academic staff, including four Professors (Tom Kilburn, Dai Edwards,
Frank Sumner and Derrick Morris).
Appendix 1: List of the principal people from Manchester University and Ferranti
Ltd. who were involved in the design and implementation of hardware and software
for the Manchester Atlas.
The following list is presented in alphabetical order. Efforts have been made to include all
relevant people but it is likely that a few names have inadvertently been omitted. The date
information (column 3) clearly requires more research. Comments are welcomed.
Name |
Dates at Manchester
University for computer
design purposes |
Dates as Ferranti
employee |
Number of
Atlas patents
on which
named |
David Aspinall |
1955 - 1970 |
|
2 |
Don Atkinson |
|
before 1961 - ?? |
|
K F (Keith) Bowden |
1958 - 1967 |
|
|
A R (Tony) Brooker |
1951 - 1967 |
|
|
John Buckle |
|
1962 - 1975 |
|
E C Y (Yao) Chen |
1957 - 1959; 1964 -1968 |
1959 - 1964 |
1 |
C R (Reg) Claber |
|
before 1961 - ?? |
|
John Clegg |
|
before 1961 - ?? |
|
A B (Ben) Cooper |
|
before 1961 - ?? |
|
Jim Doughty |
|
before 1961 - ?? |
|
Peter Duncanson |
|
before 1961 - ?? |
|
Eric Dunstan |
1953 - 1964? |
|
|
D B G (Dai) Edwards |
1948 - 1988 |
|
9 |
R L (Dick) Grimsdale |
1951 - 1961 |
|
|
Gordon Haley |
|
before 1961 - ?? |
|
P D (Peter) Hall |
|
before 1950 - ?? |
|
Brian Hardisty |
|
1961 - 1977 |
|
Douglas Hogg |
|
before 1961 - ?? |
|
D J (David) Howarth |
|
1960 - 1970 |
|
J K (Keith) Howker |
|
1959 -?? |
|
Peter Jones |
|
1961 - 1963 |
|
R H (Robin) Kerr |
1956 - 1959 |
1959 - 1964 |
|
Tom Kilburn |
1946 - 1981 |
|
14 |
P I (Peter) King |
|
before 1961 - ?? |
|
D J (Dave) Kinniment |
1962 - 1979 |
|
|
Ron Lane |
|
before 1961 - ?? |
|
M J (Mike) Lanigan |
1955 - 1962 |
|
2 |
S H (Simon) Lavington |
1962 - 1986 |
|
|
Keith Lonsdale |
|
before 1950 - ?? |
|
I R (Iain) MacCallum |
1961 - 1962 |
1962 - 1963 |
|
Frank McAulay |
|
before 1960 - ?? |
|
J J Moore |
|
before 1961 - ?? |
|
Derrick Morris |
1958 - 1980? |
|
|
Peter Mumford |
|
before 1961 - ?? |
|
|
R B (Brian) Napper |
1960 - 1997 |
|
|
S P (Phil) Patience |
1958 - 1960 |
1960 - 1963 |
|
R B (Bruce) Payne |
1956 - 1964 |
|
|
A J (Paddy) Podesta |
|
before 1960 - ?? |
|
A A (Alec) Robinson |
1948 -1950 |
1950 - 1963 ?? |
|
J S (Jeff) Rohl |
1960 - 1967 ? |
|
|
Dave Smith |
|
before 1961 - ?? |
|
John Standeven |
1961 - 1970 ? |
|
|
F H (Frank) Sumner |
1957 - 2000 ? |
|
2 |
Eric Sunderland |
|
1963 -?? |
|
Gareth Thomas |
|
1956 - ?? |
|
E T (Ianto) Warburton |
|
before 1950 - ?? |
|
Peter Warn |
|
before 1961 - ?? |
|
P B (Pete) Whitehead |
|
1963 -?? |
|
M T (Mike) Wyld |
|
before 1961 - ?? |
|
Appendix 2. Difficulties with the original MD5 magnetic drums.
This information is based on comments provided by Eric Dunstan, in an e-mail to Simon
Lavington dated 25/10/2012. Eric worked on the Atlas drums and the drum co-ordinator
from about 1957 onwards. Eric writes:
"The MD 5 drum proposed for Atlas was an upgrade of a Ferranti MD 4 drum built for
Orion but running at [as I remember 5000 rpm] which would yield a rev time of 12
milliseconds or an average access time of 6 milliseconds. The drum was designed at
Ferranti and consisted of a cylinder with a single web in the middle rotated by a motor on
top. I can't remember whether it was a DC motor or an AC motor running on a higher
frequency AC supply to get 5000 rpm.
"The outer shell [the rigid container housing the drum] had multiple slots which received
Block Heads with 16 + spares? per block. These were mounted to the shell using spring
clips and referenced to the shell via 3 'V' blocks mounted to the shell and 3 "differential"
screws with spherical hardened surfaces. The screws had LHS and RHS thread
arrangement with different thread pitches to give a "fine" adjustment. The actual system
required 25 heads in parallel (I seem to remember that it was 1/2 an Atlas word -25bits)
which required spreading the heads over more than 1 block. Since the system was
a fixed head system operating about .001" from the surface, there was a measurement
system consisting of 3 "pitot" tubes adjacent to the screws. The tubes were connected to
three manometer tubes which recorded the back pressure and as they were calibrated
gave a reading of the head media separation. There was a complication caused by the two
lower tubes being necessarily off axis and therefore suffered increased pressure on the
approach side and lowered back pressure on the retreating side of the vertical axis. This
had to be "allowed for' in the calibration.
"The whole arrangement had to be air conditioned for thermal stability and cleanliness.
The drum surface was sprayed with oxide at Ferranti but the arrangement was very fragile
and Phil Patience and I spent many nights adjusting and readjusting the separation -sometimes
the head touched the surface which we detected by a nasty smell coming out
of the drumshell. Part of the problem was the fact that the web in the middle caused the
cylinder to become "conical" because of unequal centrifugal expansion. Ferranti also had a
problem with quality control of the primer / oxide coating and I remember getting a call one
night from someone on the shift and they asked whether the surface of the drum was
supposed to be silver or brown -- it turned out that there was a problem with the primer
adherence and the whole coating had come off and was lying on the floor of the drumshell!
"At this point, it was decided to turn to a drum manufactured. in the US by Bryant
Computer Products in Vermont USA, later Excello Corporation. This was a plated drum
which was tapered and fitted with an "autolift" system which raised the whole spinning
drum. Because of the taper [narrowing to the top], the drum surface approached the pre-
adjusted individual "flying" heads which then flew at the required separation. When the
drum was "shut down" the mechanism [rather like a flyball governor] lowered whilst the
heads were still flying so that there was never head media contact. I went to the US with
Douglas Hogg to run acceptance tests on the drums, four of which were installed in the
Manchester Atlas".
References.
1. The Earl of Halsbury, Ten years of computer development. Computer Journal, vol. 1,
issue 4, January 1959, pages 153 - 159.
2. B B Swann, A history of the Ferranti Computer Department. Privately 98-page typed
document, privately circulated. (See National Archive for the History of Computing
(NAHC) catalogue reference FER/C30).
3. P Drath, The relationship between science and technology: university research and the
computer industry 1945 - 1962. Ph.D. thesis, University of Manchester, October 1973. (In
addition to studying all relevant NRDC documents, Drath was able to conduct personal
interviews with many of the key players).
4. John Hendry, Innovating for Failure: government policy and the early British computer
industry. MIT Press, 1990. ISBN: 0-262-08187-3. ( Hendry was able to study relevant
UKAEA papers that were unavailable to Drath).
5. T Kilburn, D B G Edwards, D Aspinall, Parallel addition in digital computers: a new fast
'carry' circuit. Published in Proc IEE Vol 106 Part B No. 29 Sept 1959, pages 464 - 466.
This is the first paper on the Atlas adder. The paper was presented at the IEE February
1959 Specialist Discussion Meetings in London on new digital computer techniques:
Session 6: special aspects of logical design II.
6. T Kilburn, MUSE. Proceedings of the International Conference on Information
Processing, UNESCO, Paris, 15th - 20th June 1959, page 433.
7. C Strachey, Time sharing in large fast computers. Proceedings of the International
Conference on Information Processing, UNESCO, Paris, 15th - 20th June 1959, pages 336
-341.
Refs 8 to 21 give the main journal papers describing the Atlas 1 hardware and
software, divided for convenience into sections.
The overall system and Virtual Memory:
8. Kilburn, T, Edwards, D B G, Lanigan, M J and Sumner, F H, One level storage system.
IRE Trans on Electronic Computers, Vol. EC-11, No. 2, April 1962, pages 223 - 235.
Atlas storage details:
9. Kilburn, T and Grimsdale, R L, A digital computer store with very short read time. Proc.
IEE, Vol. 107, Part B, No. 36, Nov. 1960, pages 567 - 572.
10. Edwards, D B G, Lanigan, M J and Kilburn, T, Ferrite core memory systems with rapid
cycle times. Proc. IEE, Vol. 107, Part B, No. 36, Nov. 1960, pages 585 - 598.
11. Edwards, D B G, Aspinall, D and Lane, R, Addressing magnetic tape on the Atlas 1
computing machine. Computer Journal, Vol. 8, No. 4, Jan. 1966, pages 323 - 329.
Atlas arithmetic details:
12. Kilburn, T, Edwards, D B G and Aspinall, D, A parallel arithmetic unit using a saturated
transistor fast-carry circuit. Proc. IEE, Vol. 107, Part B, No. 36, Nov. 1960, pages 573 -
584.
13. Sumner, F H, Haley, G and Chen, E C Y, The central control unit of the Atlas
computer. Proc. IFIP Congress, 1962, pages 657 - 663.
Atlas Supervisor details:
14. Kilburn, T, Howarth, D J, Payne, R B and Sumner, F H, The Manchester University
Atlas operating system: Part 1, internal organisation. Computer Journal, Vol. 4, No. 3,
1961, pages 222 - 225. Part 2, users description. Computer Journal, Vol. 4, No. 3, 1961,
pages 226 - 229.
15. Kilburn, T, Payne, R B and Howarth, D J, The Atlas Supervisor. AFIPS Proc. Eastern
Joint Computer Conference, vol. 20, pp. 279-294, December 1961.
16. Howarth, D J, Jones, P D and Wyld, M T, The Atlas Scheduling System. Computer
Journal, Vol. 5 No. 3, November 1962, pages 238 - 244.
17. Morris, D, Sumner, F H and Wyld, M T, An appraisal of the Atlas Supervisor. Proc.
ACM National meeting, 1967, pages 67 - 75.
Atlas Compiler and language details:
18. Brooker R A , MacCallum I R, Morris D and Rohl J S, The Compiler Compiler. Annual
Review of Automatic Programming Vol 3, 1963, page 229. Published by Pergamon,
London.
19. Brooker, R A, Rohl, J S and Clark, S R, The main features of the Atlas Autocode.
Computer Journal, Vol. 8, 1965, pages 303 - 310.
20. Morris, D and Rohl, J S, The Atlas compiler system. Computer Journal, Vol. 10, 1967,
pages 227 - 230.
21. Brooker, R A, Morris, D and Rohl, J S, Experience with the Compiler Compiler.
Computer Journal, vol. 9, number 4, February 1967, pages 345 - 349, 1967.
General references, continued.
22. Campbell-Kelly, M, Christopher Strachey, 1916 - 1975: a biographical note. Annals of
the History of Computing, vol. 7 number 1 January 1985, pages 19 - 42.
23. ATLAS computer Project Progress Reports. These form a group of four bound Reports
and five typescripts. The bound Reports cover the following periods: (a) up to 30th Sept.
1960; (b) up to 31st March 1961; (c) up to 30th Sept. 1961; (d) up to March 1962. The
collection is catalogued as NAHC/FER C20a, at the National Archive for the History of
Computing.
24. T Kilburn's notes on the MUSE/Atlas project, 1959 onwards. NAHC catalogue MUC
C18. This includes a handwritten sheets on package counts, 1961.
25. D B G Edwards, undated handwritten sheet of package-count calculations. The details
of peripheral equipment imply a date of this document as 1964 or 1965. See also e-mail
from DBGE to SHL dated 19th October 2012.
26. Minutes of the NIRNS Atlas Computer Committee. This committee held its first
meeting on 7th July 1961 and its ninth (and last?) meeting on 8th October 1964. The
Minutes of all nine meetings are available at:
http://www.chilton-computing.org.uk/acl/literature/minutes/nirnsaccmtgs/mtg006.htm
27. Iain MacCallum, e-mail to SHL, 28/8/2012.
28. Tony Brooker, e-mail to SHL, 12/9/2012.
29 (a). Edwards, D B G, Bowden, K F, Standeven J and Mills, O S, A Small Digital
Computer applied to Real Time Data Collection. British Computer Society Bulletin on Real
Time Systems, Vol 10, No.1 June 1966.
(b). Standeven, J, Bowden, K F and Edwards, D B G, An Operating System for a Small
Computer providing Time-Shared Computing and Control Functions. I.F.I.P. Congress,
Edinburgh, August 1968.
30. Lavington, S H and Rosenthal, L E, Some facilities for speech processing by
computer. Computer Journal, vol. 9, number 4, 1967, pages 330 - 339.
31. Riding, G, Computing Service, 1958 - 1970. This typewritten informal history, dated
12th
November 1987, was written at the request of D B G Edwards in preparation for the
40th
anniversary of the Manchester Baby computer (the SSEM) in June 1988. Copy sent
to SHL in an e-mail from D B G Edwards dated 23/10/2012.
32. Edwards, D B G, Wolff, C H, Lakhani, A H and Aspinall, D, On line Data Transmission
to the Atlas Computing Machine at Manchester University. IEE Colloquium on Data and
Telegraph Transmission, 11th Oct 1965.
33. (a). Kilburn, T, Morris, D, Rohl, J S, Sumner, F H, A System Design Proposal. Proc.
IFIP Congress 1968, Edinburgh, Section D, pages 76 - 80.
(b). Aspinall, D, Kinniment, D J, Edwards, D B G, An Integrated Associative Memory
Matrix. Proc. IFIP Congress 1968, Edinburgh, Section D, pages 80 - 90.
12th