The purchase of an AP120-B processor has been approved. It is a high-speed floating-point array processor produced by Floating Point Systems Incorporated and is intended to work in parallel with a host computer. The version chosen will have 64K words of data memory and will be fitted to the PRIME 400 computer at Chilton. It should enable calculations on large matrices to be carried out at least ten times faster; at present such calculations could only be run by letting response time increase to a totally unacceptable level.
A version of HASP is now available on the GEC 4070 at Chilton. This permits users to communicate with the 360/195s. We have also introduced an improved link between the PRIME 400 and the GEC 4070. It uses an EPSS high-level communication protocol and enables users to transfer files between the two machines. The protocol is also being used to transfer files between the PRIME 400 here and the recently upgraded PRIME 400 at Nottingham University.
Three large mini-computers will be installed in universities or polytechnics in 1978-9. Applications will be considered from those institutions which were unsuccessful in their attempts to get upgrades of existing minis, and also from those which responded to the invitation to apply for a GEC 4070 in July 1977. The minis now being offered will provide a multi-user interactive computing service in the same way as the six upgrades of existing university minis. The machines to be installed this time will be either PRIME 400s or GEC 4070s; the decision will depend on an assessment of the two machines made by Rutherford Laboratory staff in collaboration with university users. A preliminary report on the assessment is given in Section 6, below.
There are plans for a total of 15 multi-user minis to be installed in universities over the next five years, Applications have been invited from all universities and polytechnics. Of the sites where existing minis are being upgraded, Nottingham, Leeds and Sheffield are running a service on enhanced hardware. The hardware for the upgrades has been delivered and accepted at the other three sites (Swansea, Southampton and Oxford) and services are being developed.
The group has made several recommendations about software. Batch packages ASTAP, SPICE-2 and NAP-2 should be mounted and supported on the 360/195s at the Rutherford Laboratory; this has now been done for ASTAP which analyses large systems and can predict their statistical behaviour. SPICE-2 analyses transient behaviour of large systems; its modelling capacity is restricted compared to ASTAP but it is cheaper to run. NAP-2 is a research package for analysing the transient behaviour of medium-sized systems.
The following packages should be made available on the interactive computers: SIM11 for general system simulation; ITAP for transient analysis of small systems; ICAP for linear small-system analysis in the frequency domain, and SP for analysis of two-port structures.
Additionally, packages are still needed for:
The MINNIE-11 package, with its advanced interface and I/O capability, should be acquired and assessed in detail .
Experience of various APL systems is necessary to select a version suitable for general use. The one on the Edinburgh DEC10 is excellent but can make very heavy demands on machine resources.
The programme of regional meetings is almost complete. The last one will be held at Bristol on 2 May.
An open meeting for interested research workers was held at the Rutherford Laboratory on 25 January 1978.
The meeting was chaired by Professor R J A Paul. His introductory talk surveyed the work of the Special Interest Group over the past year and Mr C W Trowbridge summarised recent developments. Talks followed on particular research projects and Mr Peter Davey reviewed the work of the Interactive Computing Facility.
Interactive graphics software was demonstrated on the PRIME 400 during the day.
The meeting concluded with a discussion period. The following recommendations were made:
The Special Interest Group is now considering what action to take on these recommendations.
A Special Interest Group has been set up. The members are:
I am secretary of the group. It held its first exploratory meeting on 14 March.
A users' meeting was held at UMIST on 13 April. Details will be published later.
GAELIC is a suite of computer programs originally written to assist in the design and production of integrated circuit masks. Most of the programs are concerned with the production of two-dimensional drawings. Others cater for the more specialised requirements of integrated circuit design and for thin and thick film circuit design. The sophisticated data structure allows very large drawings to be handled efficiently on a time-sharing computer.
The programs are at present running on the DEC10 at Edinburgh. They also run on the DEC20 at Robert Gordon's Institute of Technology in Aberdeen and can be converted to run on other computers.
The programs allow the user to enter an accurate description of a drawing into the computer, to check and modify the drawing interactively and finally to produce drive tapes for artwork generation. This part of the suite does not aim to produce a layout or drawing automatically but aims to remove the tedious repetitive work from present design methods. enabling the designer to concentrate on the creative side of the drawing process.
The description of a drawing may be entered into the data structure file in three ways: by using a digitiser, by using the GAELIC input language or by drawing directly on to the screen of a Tektronix storage tube terminal.
The digitiser works from an accurate drawing on a gridded mylar sheet. The resulting tape containing a coded form of the drawing is fed into the computer and processed to give the data structure file. The GAELIC input language is usually used for smaller sketches; a file containing the input language description of the drawing is converted to give the data structure. Drawing on to the screen of the Tektronix uses the cross-hair cursor and enters the shape descriptions directly into the data structure.
The Tektronix is also used to check all or part of a drawing by plotting it out on the screen. Corners on shapes can be identified with the cross-hair cursor and either the shape or the corner moved.
The part of the drawing that is plotted can be quickly varied by 'windowing' and 'zooming' and whole areas of the drawing can be moved at once. There are facilities in the data structure for defining a series of shapes as a group; the user can call up 'instances' of this group in various positions and orientations in the drawing. It is also possible to define a matrix of patterns as a single 'instance'.
The general drawing programs have been used to design a timber-framed house and computer installation layouts, for editing musical scores and for producing flow diagrams. They can be used for any two-dimensional drawing application.
The programs for integrated circuit layout design include one which checks that the layout obeys the rules for the particular semiconductor process to be used and one which creates a logic diagram from the data structure. A third automatically places and routes a set of previously defined groups to form a complete integrated circuit. Others are used for thick and thin film circuit layout; some design the resistor geometry automatically given the value of the resistance required, the resistivity of the material and the heat dissipation constraints. Programs can also design a resistor to fit within a given rectangle. The resistor designs can have special areas for laser trimming if required.
The final programs produce drive tapes for high-precision plotters and other mask-making machines.
I am grateful to Dr John Eades of Robert Gordon's Institute of Technology for this report. Further information is available from him or from Compeda Ltd.
The DN82 network node computer at Rutherford Laboratory is now fully operational and completes a triangular network with the Edinburgh and UMIST DEC10s. Communications failures between Rutherford and either of the DEC10 sites are (apart from a slower response time) practically invisible to the DN82 users as the alternative route round the triangle may be used to access the DEC10 of their choice. The first group of users ( Cranfield, Swansea , Kingston Polytechnic, and Imperial College) have been connected to the DN82. The capacity released on the main Rutherford-Edinburgh and Rutherford-UMIST lines has been used to increase the speed of the DN82 connections to 7200 bits per second. The remaining 2400 bits per second channel on the Edinburgh line is scheduled for the DEC10 to IBM 360/195 link. The one on the UMIST line is for a connection between Leeds and the PRIME 400 at Rutherford.
There are three ports on the DN82 for remote users to dial-up. One links to the Datel 600 service and the other two link to Datel 200. Two further Datel 600 ports will be added when there is sufficient demand. The remaining four DN82 terminal ports will be connected to external user sites in the near future.
The performance of the network has been far from satisfactory due to shortcomings in the DEC software, with UMIST users suffering the bulk of the crashes; we apologise for this. DEC have given some help but solutions have had to wait until our staff gained sufficient expertise on DECNET software. Considerable progress has been made in the elimination of crashes and apart from terminals "freezing up" for no apparent reason, the reliability of the network has dramatically improved. The "frozen terminal" problem is now being actively investigated with DEC assistance.
Users of the DN82 from Tektronix terminals are the first to try a special interface unit which allows asynchronous terminals to be run at the full speed of the synchronous communications line. Previously they could only be run at one-third of the line speed. This Asynchronous/Synchronous Retiming Unit has been specially designed for the Interactive Computing Facility by Telecommunications Group of Computing & Automation Division at the Rutherford Laboratory. It will be fitted to other machines belonging to the Interactive Computing Facility in the near future.
The Technical Group Report recommended that, as part of the assessment of the multi-user minis, a collaborative project should be set up with various university and polytechnic engineering departments. University staff would mount existing running programs on both minis and report on their experiences.
Two programs were chosen. The first is a multivariable feedback design package, originally heavily overlaid. The second package is concerned with the design of shapes and surfaces.
Seven programs have been mounted:
RISP, an interactive simulation package was mounted. It analyses a description of a system as a sequence of first-order differential equations.
All groups are active research engineers, typical of those who will use the multi-user minis. The programs originally ran on a variety of different computers and operating systems.
The collaboration has been very successful. The users have cooperated fully and their views have been a great help in shaping the policy for systems programming. The PRIME 400 is capable of running a service for users in engineering, although further debug aids are needed. The GEC 4070 is also capable of running a service, but as yet we have less experience of its capacity for handling large programs.
The Technical Group report outlined a suggested configuration for multi-user minis. The following modifications to the configuration have been proposed:
A second group of collaborators has started work on the GEC 4070. They are from Salford University (Department of Mathematics), Leeds Polytechnic (School of Architecture and Landscape Design) and Glasgow University (Department of Mechanical Engineering).
There is to be a link from the UMIST DEC10 to this Unit, which is run by Manchester University. The Interactive Computing Facility has agreed to upgrade the installation substantially to support several users simultaneously. In return it will acquire a share of the resources of the Unit. The Unit is based on a PDP11/45 controlling a Vector General high-performance refreshed display system with light-pen, keyboard, function switches, potentiometer control dials and a hardware character generator. It also has a 1.5m × 1.0m high-quality CETEC flat-bed plotter and a DMAC digitising table of the same size.
Readers interested in further details should contact Dr R Hubbold, Director of the Unit.
The Interactive Computing Facility has recently advertised for staff. The permanent posts are for systems programmers and for applications programmers with expertise in finite element computations or in analogue or digital electric circuit design. These posts are graded Scientific Officer or Higher Scientific Officer. There will also be fixed-term appointments (for three years) at a higher salary for suitable applications programmers.
The new Edinburgh DEC10 user committee has now been set up and held its first meeting on March 10. The Chairman is Dr J A M Howe of Edinburgh University. This committee will be a channel for users' complaints and also for informing users about developments in the Interactive Computing Facility which affect them.
In general, Edinburgh or UMIST DEC-10 users with problems (about computing only, please!) should first contact the manager of the machine they are using. The exception is where the problems are with communications; in this case the user should contact his nearest 'node' on the communication path which failed.
If users do not get satisfaction after doing this they should take the matter up with the appropriate user committee.
Two Hewlett-Packard 2648A displays have been added to to the pool of terminals. A Tektronix 4025 display unit is to be investigated to see if it is suitable for users of the Interactive Computing Facility.
We have assessed colour raster-scan display terminals. The Interactive Computing Facilities Committee is considering the report before deciding whether or not to purchase one.
The Committee has suggested that we take advantage of the new 300 Mbyte disks on the PRIME 400 to provide an on-line information service about the Facility. As a first step the contents of 'Rapid Response' No. 5 (minus the photograph!) have been put in file NO.5 in directory RAPID-RESPONSE. By the time this issue reaches you. its contents will be in file NO.6 in the same place.
Details of the locations of the 175 terminals in the pool are kept on a file on the PRIME 400. The manual produced for users of the PRIME by the staff of the Interactive Computing Facility is also in a PRIME file, as is its index. We hope to make this information available to users shortly.
The ABACUS building design software from Strathclyde University is now running on the Edinburgh DEC10.
Some readers have wanted more copies but have not been sure how to get them. This was a surprise (people actually read 'Rapid Response'?) but the solution is easy - just let me know how many extra copies you need.
