DEUCE to KDF9 - BAC Warton - John Halliday

In 1955 Warton had some work done on the NPL PILOT ACE for the evaluation of aerodynamic derivatives and the structural analysis of the main wing box for the EECo. P1A (later to be known as the Lightning). The programmer for this work was John Dennison of the Stafford branch. The structural analysis work involved the solution of 117 simultaneous equations which was an enormous number for the time. Shortly afterwards John H. McDonnell and Philip Tattersall of the Aerodynamics department were sent on the DEUCE programming course at Stafford. The following year Alan Peacock, Gordon Pitt, and myself from the Stress office also went on the programming course. We were simply given the programming manual the week before the course and told to read it. We all found it quite a shock because the concepts were so alien at the time. Some time later Philip H. Roberts from Stress office also went on the course.

At this time we had the use of the Stafford DEUCE every Monday afternoon. We used to travel down to Stafford in the morning, have lunch at the Swan hotel, and then go out to Blackheath Lane. If you were lucky you might get three ten-minute hands on sessions on the computer during the afternoon and early evening. In this time you had to get as much information as you could from your program testing runs in order to debug your program during the rest of the week when you were back at Warton. Production runs were done by the Stafford people for us during the week and not during the Monday hands-on time. In the evening we got the London to Blackpool train at Stafford at 7:40 p.m. which was just in time for the second sitting for dinner. Sometimes if we required more time we would work on into the evening or even overnight before returning to the Swan for a few hours rest and then by train to Preston and Warton for the next day's work. In the course of these trips to Stafford we got to know most of the people on your list and also many at Whetstone and the NPL. I can endorse most of the remarks made about them on your website.

It was soon recognised in Aerodynamics dep't that programming was a new discipline and so the programmers were separated off and called Maths. Services department under John McDonnell. About the same time the first DEUCE was installed at Warton and because it was a piece of electronic equipment it became part of Electronics department. The engineer in charge was Steven Allcock and he worked for Dr. Thomas A. Duerden who was responsible to the head of department W.(Bill) Coulshed. John McDonnell was also responsible to Tom Duerden. Steven Allcock also had a Canadian assistant whose first name was Alan but I cannot recall his surname. Additional people were recruited into Maths Services and the first group included Richard Keyser, Roy M. Smith and Ronald P. Green.

Most of the work which we did in Stress Office was structural analysis which was essentially matrix algebra and solution of large numbers of simultaneous equations. The theory behind this work was led by Ian C. Taig but I think the ideas were initiated by Sydney Kelsey who had left Warton a little earlier to join Prof. J.H. Argyris at Imperial College. The work on DEUCE was done using a system called 'Scheme B'. Two matrix algebra systems had been developed for DEUCE and they were known as 'Scheme A' and 'Scheme B'. In Scheme A the matrices required for an operation were read in as they were wanted and the result matrix punched out as it was produced. This enabled the largest possible matrices to be handled but was slow and involved a lot of card handling. In Scheme B the matrices used and the result were all stored in the machine as far as possible which was much more convenient but meant that only matrices up to about 40x40 could be handled. To enable more to be stored we tended to standardise on 30x30. This meant that the large matrices we were working with had to be partitioned and all the programs written in terms of partition operations resulting in much more complexity. We had little use for Scheme A and used Scheme B exclusively. We were only able to tackle matrices as large as we did because they were sparse and banded with the result that, if one chose the partitioning carefully, many of the partitions were totally null and could be ignored. Because the names given to matrices in Scheme B were eight-bit binary numbers we became proficient at reading such numbers just as well as normal decimal. Scheme B used to indicate on the lights on the front of the machine which matrix number it required next and we had many trays of punched cards on the desk in front of us waiting to put the requested matrix in the card reader. Some jobs took many hours or even days to complete. In those days we used Finite Element methods exclusively, Force Methods were still to come.

In 1961 I moved into Maths. Services to broaden the scope of programming work I could get involved in and about the same time Gordon Pitt left the Company. Dick Keyser also departed about that time. Other people involved about this time included Ronald M. Graham, W.(Bill) Moxham, David A. Beck, Peter W. Leakey, Oliver J. Whitaker, Barry J. O'Neill, Brian A. Wardman, C. Philip Brightling, K. David Copson, W. A.(Bill)Coles, James Malloch, Alan Jennings, and Philip J. Coates( I am not sure ALL of those people got involved with DEUCE, some may have only used later machines, but certainly most of them did).There were also programmers in a number of other departments but the only name I can recall is Ron Thomas in Flight Test. By that time a second DEUCE had also been installed and I think one of the machines had been equipped with paper tape I/O.

When KDF9 was in the offing I was appointed to learn how to program that machine but there were delays and by that time our problems had outgrown DEUCE. Bill Moxham was already doing work on the IBM 7090 at the Data Centre in London using FORTRAN and Warton eventually went that way installing an IBM 7040 which almost immediately was upgraded to a 7044 and the DEUCEs were phased out.

© John Halliday - 8 July 2004