A review of the first two complete years of the Interactive Computing Facility (1977-8) was presented to the March meeting of the Science Research Council. This and the future plans were favourably received: the ICF is now on a sound footing until 1985.
There have been recent changes in the management of computing at the Rutherford Laboratory. Professor F R A Hopgood has been promoted to the post of Head of Atlas Computing Division, in succession to Dr G Manning who is to be Director of the Rutherford Laboratory following the merger of the Rutherford and Appleton Laboratories. Mr P G Davey, the Head of the Interactive Computing Facility, remains responsible for all staff working on the ICF. Mr J E Hailstone takes charge of all non-ICF work in Atlas Computing Division and will also be responsible for monitoring the use of ICF machines.
In August, Mr W Walkinshaw. Head of Computing and Automation Division is due to retire. That Division manages the two IBM 360/195s which constitute the major part of SRC's Central Computing Facility (see below, section 6 and also issue no. 5, section 6). It will merge with Atlas Computing Division on September 1st and Professor Hopgood will then be Head of the combined Division and Head of Computing Services at the Rutherford Laboratory.
The programme of work authorised by the Interactive Computing Facilities Committee (ICFC) is now well under way. This report outlines what has been done so far and the work in progress.
The THESEUS data base management system is now available in the PRIME 400 and can be used for preparing geometric and material data for the three-dimensional interactive magnet design program, GFUN3D. A user manual has been written describing release 1 of the system.
PE2D is a code for solving non-linear 2-dimensional problems, including steady state eddy currents, electrostatic, magnetostatic and current flow. Small problems can be solved in the PRIME and larger ones (up to 10,000 nodes) in the IBM 360/195. Both versions are controlled by the interactive graphics pre- and post-processor resident in the PRIME. A skeleton user guide is available.
A new boundary-integral program BIM2D for 2-dimensional linear magnetostatics is a very effective tool for computer-aided design of magnets. The algorithm is based on a total scalar potential formulation. It is available for a wide range of terminals including colour raster devices. The improved speed performance gained by using the AP120B array processor has made boundary optimisation viable. An extensive HELP facility has been implemented to assist the user.
EDDY is a 2-dimensional integral equation eddy current code which can now be submitted to the IBM system from files of data prepared in the PRIME. Movie film option using the FR80 is also available.
A new 3-dimensional code known as TOSCA. based on the use of total and reduced scalar potentials, has now been written and applied to several problems for which GFUN3D proved inadequate. It is at present resident in the IBM system but interactive pre- and post-processors will be provided in due course. The new code has been designed to solve boundary value problems and is particularly useful for investigation of electrical machines with small air gaps.
BIM3D, the 3-dimensional version of the boundary integral method program, is now available in batch mode on the IBM system. Pre- and post-processors resident in the PRIME will be provided.
Other development work includes investigating the possible use of a hybrid method for field calculations; it would have a differential formulation in interior regions of the problem and an integral formulation for external regions. This method appears to be expensive in both core and CPU time. Alternatively, 'ballooning' methods or infinite elements may be cheaper, using a differential formulation throughout. Also, an exhaustive study has been carried out of the alternative methods for solving eddy currents in 3 dimensions and a detailed report will appear later this year.
An open meeting of users will be held at Rutherford Laboratory on Friday July 6th. Readers interested in further information should contact Jim Diserens.
The ICFC has approved proposals for Electric Circuit Design software. An integrated design system is under development by the Digital and Analogue Circuit Design (DACD) group at the Rutherford Laboratory. This will rely upon a graphical interface with the designer using refreshed displays. Such a system was demonstrated some years ago by Dr Spence's group at Imperial College (MINNIE) and the falling cost of refresh graphics terminals makes this "natural" method of communication by the designer feasible on the ICF in the near future. Compatibility with the more commonly available devices (eg VDUs and teletypes) will of course be maintained.
Meanwhile work has gone ahead to provide more conventional analytic facilities for immediate use. For interactive small system analysis, Professor Nichols of Southampton University has provided ITAP, ICAP and SP. These allow DC, AC and transient analysis of conventionally specified circuits and have some facilities for two-port structures using scattering-parameter formalism. Considerable work has also been done to ease the use of the large packages (ASTAP, SPICE and NAP/ANP) on the batch machines. These can now be accessed from the ICF without the user needing to know IBM job-control language. Source files and results are handled automatically on the ICF. This service is backed by an extensive HELP and information system which is kept up to date on the PRIME 400 and GEC machines.
The complete DACD system will provide easily-used facilities for:
The ICFC has also approved the proposal to set up a microprocessor development system on the network as recommended by the Special Interest Group for Digital Design (see 'Rapid Response' No.8, section 2). The DACD group was asked to seek the comments of other major Government organisations which were contemplating microprocessor programmes before going ahead. This was to avoid overlaps in provision and to ensure some coordination of effort. Anticipating this approval, some work has been done to locate centres of microprocessor expertise in University and Polytechnic Departments and to finalise plans for a system which will provide:
This system will allow a potential user to assess quickly the relative merits of competing devices for a particular application. He can also simulate either the chosen device or the gross stimulus-response behaviour of the system to which it will be attached (or both) before embarking on hardware production. When the hardware specification is finalised the user will be able to design and "debug" the software before delivery of the hardware. This should result in substantial reduction of the equipment lead times which needlessly delay many research and development programmes.
First-time users will be helped by "worked examples" and specimen solutions to past problems with interactive access to an idealised instruction set that easily converts to specific applications.
The members of the DACD group are: Peter Dewar, Ian Benest, Robert Gay, and John Mclean.
The two Special Interest Groups (SIGs) in Finite Elements made a joint submission to ICFC. They requested manpower over the next five years to provide FE software for the research community in Universities and Polytechnics ( see 'Rapid Response' No.4, section 2 and No.7, section 5). The research community obviously has a very wide spectrum of requirements in the FE area, but divides into two major groups the 'users' and the 'developers'. The first group are essentially problem-oriented: the need is to, say, analyse a specific structure under specific loadings to determine the stress and displacement fields. The requirement is for a program which can carry out the required analysis, is easy to use and reliable and has good interactive pre- and post-processing facilities. The second group the developers - are interested in new types of analysis: details of the program are important and access to source code is usually necessary.
These two groups of users obviously require quite different forms of software. The SIGs decided that the 'user' group requirements could best be met by supplying developed commercial-standard packages which could be used as 'black boxes' where necessary. For the 'developers', it was decided that a library of FE software, modelled on the lines of the NAG library, was the best solution. A Steering Group has been set up to oversee development of the facilities over the next few years. The Steering Group members are:
Four packages have been under evaluation at the Rutherford Laboratory for the past year. A number of test problems have been devised and solved, in so far as was possible, on the four packages. This phase is now complete and a meeting of the "structural applications" SIG will be held shortly. The intention is that two of the four packages will be selected for support.
The FE library is now under development by staff at Rutherford, with the active help of the Steering Group members . The library is at two levels: Level 0, the set of basic building blocks, and Level 1, a set of example programs showing how the Level 0 library can be used to solve a wide range of problems. Approximately half the effort will be devoted to 'structural' engineering in its widest sense. The other half will be invested in the use of the FE technique to solve a range of problems in fluid mechanics; in essence a demonstration of the use of finite elements as a partial differential equation solver. There will be continuous development of the FE library over the next few years under the guidance of the FE Steering Group. The initial release is scheduled for the end of September this year, complete with documentation which fully describes both library levels.
The staff carrying out the work described above are: Ken Robinson, Subodh Chanda, Chris Greenough, and Tony Ing.
A very necessary requirement for FE software is pre- and post-processing, preferably interactive with graphics. An ICF Working Party is currently considering this problem (see 'Rapid Response' No.8. section 6). A commercial pre-processor FEMGEN, which is capable of being interfaced to almost any FE system, is being purchased for evaluation on the PRIME 400 computer at Rutherford. If found suitable, the package will be made available on those ICF machines with users requiring such a facility. It provides good facilities to aid model specification and mesh generation, and will be an adequate stop-gap until the results of the Working Party's work are available.
A full description of the facilities available, how to get access to them, and a chance to air your own views will be given later this year at the next FE User Meeting. This is scheduled for Tuesday September 18th at the Rutherford Laboratory; anyone interested in this meeting, or anyone who would like to comment or make enquiries before the meeting, should contact Ken Robinson at Rutherford.
Installation of GEC 4070 multi-user minis (MUMS) in universities etc has gone far better than expected in most respects and disastrously in one: it has taken much longer than expected to complete links to the Rutherford Laboratory even when all that was lacking (as at Cambridge or Newcastle) was a relatively short line. The machines now installed are at Bristol, Cambridge, Glasgow, Cranfield and Newcastle. Bristol, Cranfield and Glasgow now have links to Rutherford and are connected to each other, to the Rutherford 4070 and to the 360/195s via the SRC X-25 network.
Installation of the machines has been perfectly smooth. In all cases it has taken 3 or at most 4 working days. The first day involves physically easing the machine through the university labyrinth and connecting it to the power supplies. The second day is spent tediously observing acceptance tests and on the third day the system is mounted. Thereafter the users take over. It is also a pleasure to record that reliability of both hardware and software has been good with a few niggling and not very serious exceptions. The philosophy of supporting the machines from Rutherford is remarkably successful considering the lack of physical communications links. We are confident that when these links become available we will have a sound and economic system.
Future GEC installations will be at Bradford, Cardiff and (almost certainly) Birmingham. Deliveries will be at the end of the year.
Among the other multi-user mini sites selected in the last round, Sussex University is to receive a PRIME 550; this is the result of a submission made jointly by the University and Brighton Polytechnic. A decision will shortly be made about which type of machine is to be installed at Surrey University.
The installation of a total of 10 new MUMS in universities or polytechnics has now been approved, including those mentioned above. ICF plans include the selection of up to five further sites by the beginning of next year.
A PRIME 550 is to be purchased for the City University, London. The main reason is to support Professor L Finkelstein's Instrument System Centre; the equipment will be paid for by Engineering Board's Control Engineering Committee but installed and supported by ICF.
An open day was held on 14th March 1979 to enable friends and colleagues to see some of the work being carried out by users of the PDP11/70. Demonstrations included:
The open day aroused considerable interest in the ICF. The Applied Science Group at Leeds, which is responsible for the PDP11/70, publishes an occasional newsletter. Anyone wishing to be added to the mailing list should contact Veronica Boulton.
The ICF PRIME 400 at Rutherford has been upgraded by the addition of a second processor with 1 Mbyte memory. This upgrade was supported by the Engineering Board's Electrical and Systems Engineering Committee owing to the very heavy load on ICF equipment that is being generated by potential users of the Electron Beam Lithography facility. All hardware is installed except the 'PRIMENET Node Controller', a high speed coaxial link between the processors. At present the machines are connected by a 9.6 kbaud synchronous line. PRIMENET allows the two machines to be run effectively as one, although users will be generally allocated usernames on only one machine. Despite the slow link, it has been necessary to begin to move users on to the new machine ('PRIME B') to relieve the load on the original 'PRIME A'.
An IBM 3032 has been added to the IBM 360/195 complex at the Rutherford Laboratory. It has been installed to take on the front-ending work currently carried out on the two existing coupled 360/195 computers which provide batch processing facilities.
It is necessary to develop the software enabling the 3032 to manage the large number of workstations and terminals, and so the 3032 has been attached initially as a further back-end processor on the 360/195 complex. In this mode, it will provide extra batch processing capacity and scope for an increase from 60 to 70 active ELECTRIC users, while allowing facilities for software development for its future role which will be introduced in stages during 1979 and the early part of 1980.
Further information about the SRC batch processing facilities may be obtained from Mr E B Fossey, Head of User Support Group, Computing and Automation Division, Rutherford Laboratory.
The current discussions about the future of the DEC-10 installations at Edinburgh and UMIST are aimed at establishing and clarifying the characteristics and future needs of the user population associated with these computer systems. The ICFC is therefore taking a closer look at some of the statistical material available, in an attempt to identify features which can be used in providing future facilities. Mr A Payne from the University of Waikato (New Zealand), who is on a sabbatical visit to the UK, has agreed to examine this study of the user profile and to report during the summer. This is one of a number of different studies being carried out; users of the DEC-10s are also being asked to make their views about the effectiveness of the machines in their research known to their local User Committees. It is expected that the ICFC will consider the information during the coming autumn when forward look estimates for 1980/86 are prepared.
The last issue, section 5, stated that on ICF machines all batch jobs attract the "discount" charging rate for machine accounting, irrespective of the time of day or week at which they were run. This is not true for the UMIST DEC-10, where batch use will continue to be charged at the rate appropriate to the time-band during which the job is run. In the same section of the last issue it was also stated that "all ICF have rationing schemes." I have been asked to point out that the Edinburgh resource control system is not a rationing scheme.
Professor Hopgood participated in the 2nd IFIP Workshop on the Methodology of Interaction at Seillac, France. This workshop, known as Seillac II, covered the Methodology of Interaction. Seillac I in 1976 played a major part in getting progress restarted towards effective standardisation in computer graphics, after several years when all efforts seemed abortive. It is now clear that lack of understanding of man-machines interaction is one of the major remaining obstacles to the definition of effective graphics standards, which is why Seillac II was arranged. Its work should be as effective and useful as that of Seilllac I, and if so it will make a substantial contribution to the future of interactive computing. It is hoped that the results of the workshop will be published in September.
