David John Evans was born on 30 September 1928 in Pantyffynon, a small hamlet in Carmarthenshire, just north of Llanelli in Wales. His father worked as a train driver on the Llanelli and Mynydd Mawr Railway, and his mother in the local tin-plating industry. The family were Baptists in religion, and the virtues of industry, trustworthiness and thrift were instilled in David at an early age. Although he was destined to live in England for most of his life, David never forgot his Llanelli roots and he remained a passionate and exuberant Welshman.
Times were hard in West Wales in the 1930s and 1940s, but David was fortunate to obtain an excellent education at Llanelli Grammar School. Here, his ability in mathematics and physics became evident, and he was also a keen violinist. David became the first member of his family to attend university, going up to University College, Aberystwyth, Wales, in October 1946. He graduated three years later with a B.Sc. in Mathematics, specializing in the applied side of the subject, particularly in the disciplines of hydrodynamics and aerodynamics. It was at Aberystwyth too that he met his future wife, Naldera (Derry) Owen. They were married in 1955 and had three children.
After National Service, David pursued his interests in hydrodynamics, taking an M.Sc. in Engineering from the University of Southampton in 1955 and then joining Rolls-Royce as a senior mathematician in their research department at Derby. Although he found the mix of mathematical, industrial and experimental work at Derby congenial, David was gradually becoming disillusioned with the constraints of “pen and paper mathematics”. He was ready to welcome the coming computer revolution.
Resigning from Rolls-Royce in 1958, David became a Research Fellow of the Computing Machine Laboratory at the University of Manchester. He received his Ph.D. in 1960 for a thesis entitled Iterative methods for solving linear systems. In 1964, David was offered a readership and the directorship of the Computing Laboratory at the University of Sheffield. Here, he wrote influential early papers on the subject of pre-conditioning. The crucial step in the successful computer solution of elliptic problems is often the pre-conditioning, or the choice of matrices, for the linear system in which the iteration takes place. The conflicting demands of solvability and improved convergence for pre-conditioners fascinated David. He repeatedly returned to the subject of pre-conditioning in his researches, finally summarizing his achievements in two books on the subject.
David moved to the inaugural Chair of Computer Studies at Loughborough University in 1972. Over the next 25 years, his lively and rumbustious personality helped build up the department into one of the foremost in the United Kingdom. During this time, he supervised over 80 graduate students, as his research blossomed into new areas of numerical analysis — integration methods for ordinary and partial differential equations, neural networks, computational matrix methods, wavelets, and sparsity and its applications. He organized a number of major international conferences on computer mathematics at Loughborough, and was in regular demand as a lively and learned speaker himself. His increasing academic prominence was recognized by the award of an honorary doctorate by the University of Aberystwyth in 1974.
During the 1980s, David began to move into new fields, becoming interested in the idea of parallelism. Together with students, he designed and implemented an actual, low-cost parallel system at Loughborough to study all aspects of the emerging discipline of parallel processing. The four-processor machine built by his research team served as a test bed for many parallel versions of numerical algorithms, particularly related to the solution of partial differential equations. David's insight was that parallelism could be used to implement computationally-complex methods to overcome shortcomings in numerical techniques: for example, the group explicit methods, recursive decoupling and, of course, pre-conditioning. His interest in parallel processing led to the formation of the Parallel Algorithms Research Centre at Loughborough University in 1989.
Also in the 1980s, David became excited about unleashing the potential of massively-parallel machines using algorithmically specialized circuits. Aimed initially at chip-level, very large-scale integration, the principles of regular computation and the automatic synthesis of parallel algorithms became a popular paradigm. Throughout the 1980s and 1990s, David made contributions to both the application of systolic, regular, and other emerging computational domains. Using his vast knowledge of practical algorithms, and particularly numerical algorithms, he was able to map a range of key applications into the new paradigm.
Now, of course, there are commercially available parallel architectures, massively parallel distributed networks and cluster machines. New dual core architectures, pipelined processors, and very long instruction word computers are all available to support the scientific community in e-science, e-learning, and GRID computing. What once was a research concept is now firmly established in the world of science. Machines that David could only dream of when he began with that four-processor machine are now readily available. The algorithms and techniques David developed over his long and distinguished career have had no small part to play.
David wrote over 500 scientific papers and edited ten books, the most important of which were Parallel Processing Systems, Sparsity and Its Applications, Systolic Algorithms and Pre-conditioning Methods. His influence lives on in his many research students. David communicated to them his joy in solving problems on computers and inspired them with his encyclopaedic knowledge of numerical mathematics. He formed particularly close ties with the mathematical communities in Greece, China and Malaysia. Many of his former students now hold senior academic positions in these countries and through them David has played an important role in fostering international academic excellence.
Not the least of David's contributions was his influential editorship of a number of scientific journals, in particular Parallel Computing and Parallel Algorithms and Applications. He was Editor-in-Chief of International Journal of Computer Mathematics from 1979 until increasing ill-health forced him to resign a few months before his death on 5 October 2005.
WYN EVANS
University of Cambridge, UK
Email: [email protected]
GRAHAM MEGSON
University of Reading, UK
Email: [email protected]