The Committee, often referred to as the ICFC, met on 16 September. It considered applications for finance to upgrade more university mini-computers; these proposals were invited in the special notice sent out with the last issue of 'Rapid Response'. The proposals accepted were for the upgrade of a PDP11/45 in the Department of Engineering Science, Oxford University and of a PDP11/40 in the Department of Civil Engineering, University College Swansea. The proposals for siting a second multi-user mini in a university will not be considered in detail until further discussion has taken place.
There are some minor corrections to section 4 of the last issue ('Accounting for computer resources') :
The functions of the Special Interest Groups(SIGs) are:
We would welcome comments on the reports below or on any other relevant subjects, or reports of experience with suitable software.
Both the SIGs concerned with the applications of finite element methods met in July. The name of one of them Building Analysis was changed to Structural and Geo-Technical Applications, or SIGSG. The Fluid Mechanics group were content with their name and will be referred to as SIGFM. The secretary of both groups is Ken Robinson at the Rutherford Laboratory.
In the short term user needs could best be met by a program package capable of linear static and dynamic analyses (the latter at least sufficient for eigenvalue extraction). The elements provided should include: plane stress, plane strain, plate bending, 3-D plates, frame elements, shells and axisymmetric elements. There is some requirement for non-linear analyses and direct code involvement is probably necessary. Documentation and support should be to a high standard, In the longer term, the requirements could best be met by a library of finite element routines which would be centrally maintained. Source code should be available to users. The system should be capable of handling non-linear analyses or of being extended to handle them. Enhancements to the basic library could be undertaken over the five-year period.
A further meeting was held on 3 October. Details will be published later.
In this field there is no commercial or generally accepted software available. The SIG chairman, Dr C A Brebbia of Southampton University, has written to various research workers asking them for programs which could be made generally available. They were also asked to describe their research interests for the next five years.
In the longer term, the requirement is essentially similar to that of SIGSG, namely for a centrally-supported FE library. Main areas of interest identified were:
Mechanical and chemical engineers also need software tools to solve their research problems. Another meeting was held on 20 September. Details will be published later.
The first meeting of the Special Interest Group on AI accepted the aim of estimating the support needed during the next five years and adopted the additional aim of looking ahead to when the limitations of the DEC-10 make it inadequate for future AI programs.
The work to be done including advice to users could justify the employment of 3 or 4 system programmers. It was decided to ask for the two existing programmers to be retained and appointed on a longer term basis as they will be needed for at least five years. 90% of AI work on the Interactive Computing Facility uses POP2 or POP10 and this is likely to continue. POP2.1 and POP2.5 are being developed; a PDP11 version of POP2 produced by Dr Hardy at Sussex University is attracting interest from this country and abroad. It is not expected that the language will ever be static. Although POP2 is still worth developing, a substantial part of AI research involves development of new languages. This differentiates the needs of AI workers from those of groups requiring program packages. Development work on PROLOG is being done by David Warren at Edinburgh University so additional support is at present unnecessary. Help should be provided for users of LISP but no development work undertaken. Possibly EVIL should be made available on the interactive network later, but this is not necessary at the moment. The standard allocation of disk space on the Edinburgh DEC-10 is often inadequate for vision processing and other work requiring large data files.
The secretary of the SIG is Peter Kent of the Rutherford Laboratory.
The SIG on 'Analogue and Miscellaneous Electrical Circuit Design' (or 'Lumped and Continuous Circuit Design') met on 27 June. It sees its job as assessing which set of programs is suitable for mounting. It selected the following fields for detailed investigation:
The member from the Rutherford Laboratory, Peter Dewar, was to assess the performance on the batch machines of three program packages which cover some of these topics: NAP-2 from Lyngby in Denmark, SPICE-2 from Caltech and IBM's ASTAP.
There was some concern about access to the network. Is this guaranteed for all approved projects? Some such guarantee would be needed before any significant effort is put into mounting packages on the batch machines. If a guarantee is not forthcoming, it may be better to abandon the idea of a network altogether and put the Rutherford Laboratory's effort into producing and supporting a package library for mounting on university mini-computers. The members of the SIG are very concerned that the candidates selected for the Interactive Computing Facility posts at the Rutherford Laboratory should have the specialist experience needed to support such a package library. Also, it is important to make full use of the minis and terminals already installed in the universities before making extensive purchases of new equipment; already it is difficult to find out what equipment is available, where it is and whose it is. The emphasis should be on applications, particularly in industry. Perhaps people from industry should be invited to join the SIG? Since the meeting some things have become clearer. ASTAP, kindly lent by IBM for assessment, has proved very impressive; its price may be a problem . Also, any large, central analogue (or digital) modelling facility could usefully serve a wide range of non-electronic specialists. Some demands have arisen from fields such as biology, ecology/social science and mechanical simulation (eg non-linear vibration). The major computing facility a powerful electronic analysis package - would need to be 'buffered' from the users by powerful pre- and post-processing facilities, each with its bias towards particular subject areas. The researcher then does not need to draw the electrical analogues explicitly.
The second Circuit Design SIG, on Digital and, Integrated Circuit Design, met on 27 September. A report will be published later.
The SIG on Electromagnetic Applications has met twice. The first meeting considered a survey by Mike Newman of the Rutherford Laboratory on problems which would be raised by wide dissemination of software. The SIG suggested that his recommendations could be expanded into a much-needed code of practice for all software for the Interactive Computing Facility. The SIG also felt that it was important to clarify the terms on which programs might be made available for commercial use, eg through the National Research and Development Corporation.
Subsequently, a circular letter was sent out to all electrical engineering departments in universities and polytechnics. It described the Interactive Computing Facility and asked for information on available software and on applications where improved software is needed. The replies, discussed at the second meeting, revealed a number of special-purpose programs which could be made suitable for general use if some manpower were provided. Most users' needs were already known; however, some potential users who replied had not previously known about some of the existing software or about the Interactive Computing Facility.
The SIG has proposed that manpower should be employed in developing programs for eddy currents and boundary-integral methods and should also allow flexibility in handling other topics covered by SRC grant applications, such as a hybrid program for high-frequency work. Some manpower should also be devoted to user support.
It was agreed that a second open meeting should be held for all research workers in electromagnetics who are interested. It should take place as soon as possible, perhaps in November.
The secretary of the SIG'S John Collie of the Rutherford Laboratory.
The communications network currently planned for the Interactive Computing Facility is shown on the accompanying map. Some of the equipment shown has yet to be delivered, in particular the DN82 network node computer for Chilton.
The DEC-10 machines at Edinburgh and UMIST were at first connected via Chilton, over links totalling 600 miles in length. This was a very new departure - as far as we know DECNET has previously only been used to link DEC-10s when these have been on the same site. After much hard work these links are now working satisfactorily and a direct UMIST-Edinburgh link is being tested.
The present situation is unsatisfactory for several reasons:
These problems will vanish when an SRC/University network is developed. A unit set up jointly by the SRC and the Computer Board is examining the feasibility of setting up such a network. In the meantime various developments will provide some immediate help:
Although the PRIME 400 at Chilton has had access to the 360/195 for some time, the GEC 4070 has yet to be connected. The software is still under development by GEC and should be installed shortly. Unfortunately the terminals on the two machines cannot access the rest of the network and vice versa; again one looks forward to the SRC/University network to solve this problem.
The Chairman of the UMIST user committee is Mr O S Mills of Manchester University and the secretary is Mr A U Beresford of UMIST. The user representatives are:
Each member will act as a channel for requests and complaints from users in the institution shown alongside his or her name. As the number of users increases, extra representatives will be invited to join the committee. If it gets too large there will be some amalgamation of the constituencies. The management of the computing service will be represented by Dr G C Barney and Mr W Swindells. The Rutherford Laboratory will send one representative and the Edinburgh user committee may also send one.
The membership of the new Edinburgh user committee is not yet finalised.
A very complicated set of links to the DEC-10 has grown up around the Hope Park Square area of Edinburgh, mainly to serve the Artificial Intelligence research community there. Rationalisation will be based on a cheap remote terminal concentrator; a DN92 has been ordered.
Mr Aart Bijl of the Department Architecture, Edinburgh University writes:
"In your June issue of Rapid Response. I note that the list of Special Interest Groups (SIGs) excludes the general topic of interactive graphics and, in particular, CAD applied to building. Some time ago, when SIGs were first inaugurated. I had discussions with Rutherford and it was suggested that graphics would form a special interest, but this has not materialised. This omission seems odd, in relation to a service which is intended to serve a community of CAD users and researchers. I think you will find considerable experience of interactive graphics among researchers in this field who are now on or are due to come on to the network. At EdCAAD we have concentrated a great deal of effort on this field applied to building, and we have developed and tested a system called TIGER (Terminal Interactive Graphics Executive Routines) which uses a satellite processor linked to the Edinburgh DEC-10. If Rutherford intend to come into this field, I should think it could benefit from existing experience."
Mr Bijl enclosed with his letter a brief two-page outline of the TIGER system. He will supply details to any interested reader. A discussion meeting on Computer-Aided Architectural Design was held on 23 September, chaired by Mr R John Lansdown of Turner, Lansdown, Holt and Paterson. I hope to include a report in the next issue.
The inclusion of the communications map has left no room for any more. Next issue: December.
