Manchester Computer Architectures, 1948-1975
Original Publication Date: 1993-Sep-30
Included in the Prior Art Database: 2005-Oct-07
Software Patent Institute
SIMON H. LAVINGTON: AUTHOR [+1]
The overall characteristics of the three computers under review are summarized using modern terminology in Table 2. Although the Atlas and MUS were rated among the fastest in the world at the time, the performances are today seen as quite modest. Indeed, the small amount of available RAM for each computer seems scarcely credible today. The Mark I described in Table 2 is the April 1949 version, rather than its June 1948 predecessor, which was even smaller. The so-called baby machine that first ran a program on Monday, June 21, 1948, had just 32 words of 32 bits each for its main memory, no index registers, manual input/output, and a simple seven-function instruction set. The 1949 version of the Mark I given in Table 2 had two B-lines (general- purpose index registers), 128 words of primary memory, and 1K words of drum backing memory. The production Mark I, first delivered in February 1951, extended the capacity to eight B-lines, 256 words of primary memory, and 3.75K words of drum storage. The Atlas described in Table 2 is the first production version as inaugurated at Manchester in December 1962; subsequent production Atlases had more RAM. The MUS is the one-off machine as it existed at Manchester in 1975. Note that, especially in the case of the MUS, it is somewhat misleading to give a single value for some of the parameters in Table 2. For example, as is shown later, MUS instructions could be 16,32,48, or 80 bits long. [Figure containing following caption omitted: Table 1. Summary of five Manchester computer projects. Asterisks denote qualifying comments in the text.] [Figure containing following caption omitted: Table 2. Overall characteristics of three Manchester University computer designs, expressed in modern terminology. (The production Mark I had eight general-purpose registers.)] Each machine in Table 2 occupied a floor area equivalent to a large room, putting it in the modern category of ";large mainframe."; Figures 1 through 3 show the scale of each. The technology of each machine is now only of specialized interest, but for completeness highlights are given in the sidebar on page 54. As always, however, there is no substitute for reading the original papers; the references cited here3-l7 give a selection of technical papers for the Mark I, Atlas, and MUS. A modern explanation of the Mark I and Atlas instruction sets will be found in my paper on these machines.l8 Atlas is discussed by Bell and Newelll9 and in the 1962 paper by Kilburn et al. S Bell and Newell introduce Atlas as ";one of the most important machines described in this book"; and comment that ";Atlas was about the earliest computer to be designed with a software operating system and the idea of user machine in mind."; The MUS hardware and software are covered in considerable depth in the book by Morris and Ibbett,20 and a readily accessible summary will be found in the paper by Ibbett and Capon.21 The architecture of the ICL 2900 series is described by Buckle.22 While there are some fundamental differences, the architecture of the 2900 series owes much to, and has a great deal in common with, the MUS.I
THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.
Copyright ©; 1993 by the Institute of Electrical and Electronics Engineers, Inc. All rights reserved. Used with permission.
Manchester Computer Architectures, 1948-1975
SIMON H. LAVINGTON
Because of changes in computer technology and terminology, it is often difficult for present-day observers to judge the significance of early digital computer projects. This article follows some architectural themes of interest, as they evolved in the design of three innovative Manchester University computers: the Mark I, Atlas, and MU5. It traces themes such as operand address generation, instruction formats, and memory management during the period from 1948 to 1975. These themes are illustrated by a set of normalized diagrams that may aid further study of the original references.
The purpose of this review is to explain in modern terminology the structure of the principal digital computer designs to emerge from Manchester University in the period from 1948 to 1975. There are three aims in this analysis:
1. To examine the contemporary reasons for the introduction of various features that now appear curious or novel. 2. To compare these features with contemporary developments in other centers of innovation. 3. To assess the longer-term impact, where it exists, of early Manchester ideas whose influence might be detected on computer architectures of the 1980s and 1990s.
The computer design group at the University of Manchester was responsible for five prototype machines during the period under review.! Table 1 summarizes the names of these prototypes and their UK industrial derivatives.
This article concentrates on the first, fourth, and fifth projects in Table 1 because these are the most interesting architecturally. For completeness, however, the other two machines are briefly described in the sidebar on page 53.
Before analyzing the architectures of the three computers of principal interest, it is perhaps appropriate to hint at the motivations, personalities, resources, and working environment that characterized the computer design group at Manchester University during the period under discussion. The full story will be found in my books but a quick sketch is sufficient to suggest where priorities lay.
From 1947 to about 1952 the Manchester group consisted of two or three key faculty (or equivalent), supported by an equal number of PhD students. They worked with war-surplus components and an enthusiasm for electronic innovation that was inspired by Professor F.C. Williams' notable successes in wartime radar technology a few years earlier. (See also the article by Mary Croarken in this issue.) From about 1952 to l9S9, there were about four key people, who had by then developed links with the nearby Manchester electrical engineering company Ferranti. These links provided technical support in the form of components and facilities. During this period, systems software and high-level la...