The Contribution of MU5 to the ICL VME2900 Series

Two papers on MU5 have recently been published, in the IEEE Annals of the History of Computing, Vol. 21, No. 1, January-March 1999:

Note that these are only readable on the WWW if you are a member of the IEEE Computer Society with an online subscription and an e-account. These papers are well related to the talks given by the authors at the Golden Anniversary Conference in June 1998.

Brief summaries of the MU5 project can be found on these pages as part of Tom Kilburn's biography and in the Virtual Museum.

The following summary of the contribution of MU5 to the ICL VME2900 series is derived from Roland Ibbett's paper and the talk "Continuing the Manchester Connection" given by Brian Procter, Chief Architect ICL High Performance Systems, at the Golden Anniversary Conference. The annotated slides of this talk are given on these pages (in pdf format); extended direct quotes from them are given in italics, and corresponding slide/page numbers in square brackets (use the pdf Page facility at bottom left to get to the correct page).

Work on MU5 started in 1966, once the Department of Computer Science had settled down after the production of Atlas and the creation of the new department with its new undergraduate school. With four years of experience of Atlas acquired, and with the majority of programs being written in high-level languages, the main focus of MU5 was to provide an architecture geared to the efficient running of such programs. It was observed that although Atlas had a very rich instruction set (around 400 including extracodes), relatively few were used in practice by the compiler-writers. Conversely higher-level constructs that were commonly required by the compiler-writer could be implemented at a high level by hardware, with a resulting increase in execution speed and decrease in compiled code (e.g. routine calls and array/record access to a wide range of word sizes).

The MU5 project was to produce a design for a range of compatible machines together with the associated operating systems and compilers, such as might be provided by a commercial manufacturer. However there was no prior agreement that the design would be taken up by a commercial manufacturer. Only one example of the range would be actually built (at the higher end), but the software was designed from the start to be suitable for the whole range, and indeed portable over different architectures.

[p.9] In 1966 the University approached ICT to enlist support for the MU5 project. In 1967 the University applied to the Science Research Council for a grant. In support of the application, ICT agreed to construct a specially modified version of one of their 1900 series machines (a 1905E) which would be used as a multi-processor with the new machine. ICT would provide construction facilities for the new machine at cost and would make available the technology and Design Automation tools used in its recently released top-of-the-range 1906A. ICL staff were seconded to the University to work on the project (5 out of a team of 20 in 1968). Note that ICT had now become part of the newly created ICL [p.6]. The 1905E was in fact purchased right at the beginning of the project, for use in the development of both software and hardware (with extensive use of simulation); then when the new machine was built it became the main peripheral processor of the MU5 complex.

[p.7] Around 1969 ICL started to consider the design of a New Range of computers to succeed the successful 1900 series. A set of teams was set up to formulate a set of proposals, for example basing it on an existing or proposed system, with main candidates the existing ICL 1900 series and EELM System/4 series, the Basic Language Machine, an elegant, fundamental re-think of computer architecture developed by a team in ICL's labs under the direction of John Iliffe, and MU5, now 3 years into its design.

However the proposal finally decided upon was the "Synthetic Option". This was to draw on the best modern ideas from anywhere, and to synthesise them into a coherent and flexible architecture.

[p.8] The main design influences came from the four sources indicated:-

The influence of the University work on the Synthetic Option continued the tradition started with Ferranti. The six members of the SO team were familiar with the work on MU5 and included a member of faculty working as a consultant.

The SO team shared the University view on a number of important aims and the resultant MU5 architectural decisions were well received by the SO team. Two such areas were the ideas of generalising the OS whilst improving its efficiency; and the recognition of the growing importance of High Level Languages with their need for a "compiler friendly" architecture. Although the MU5 and New Range architectures are not the same, there is a clear family resemblance. Many of the fundamental approaches to storage management and process structure in New Range are directly derived from MU5.

[p.9] During the development period of the first 2900 systems, serious consideration was given to the idea of ICL manufacturing the MU5 design as an early 2900 product. Some aspects of the two architectures were sufficiently different to cause user incompatibility so, in 1971, a convergence exercise was mounted and changes introduced to bring them closer. In the end, the idea of using MU5 in this way was not taken up.

[p.10] Eventually, in April 1974, the New Range was launched as the 2900 Series. The launch was a slick marketing affair. Regrettably, as a matter of historical fact, it must be recorded that no credit was given at the launch to the contribution made by the University.

[p.10,11] Brian Procter then went on to a brief summary of the evolution of the VME2900 series into System 39 (launched in 1985) and the Trimetra (1997). Both of these systems preserved the 2900 architectural model (in the Trimetra by emulation).

In his conclusion he was kind enough to say [p.14]:

We offer our warmest congratulations to the University of Manchester and the Department of Computer Science on the proud occasion of the 50th Anniversary of the Baby.

ICL, and particularly the Manchester-based division, High Performance Systems, is proud of its long association with the University. We trace the connection back through the predecessor companies to the pioneering days of the University and Ferranti described by earlier speakers.

We are pleased to honour not only the team and the event of 50 years ago, but the people and the stream of work which has continued to emerge from the University. We know that your work and your people are world-class. We know it because you have helped us to produce products which have stood the test of time and have held their own on the world stage against competition from the best.

Copyright The University of Manchester 1998, 1999