In March Dr.Manning resigned as Atlas Computing Division Head, and Prof F.R.A.Hopgood was appointed with effect from 1st April,1979. The Division then had 44 scientific staff members, divided into 2 groups as shown in the diagram. Mr.P.G.Davey is in charge of the Interactive Computing Facility, Mr.J.E.Hailstone looks after general computing projects. During the quarter Mr.P.C.Burkimsher left, and we welcomed two new members. Dr.R.Gay joined the Digital and Analogue Circuit Design section, and Mr.T.W.G.Dunwoody the Grants and Awards section.
I was asked to organise the experimental production of a quarterly journal for the Division, with computer assistance. Several methods were proposed, the one chosen was to request each member of the Division to enter a simple file containing the text of their journal, into a computer. The text was merely appended together to form the final journal, using RUNOFF on the PRIME400 computer, with output printed on the Diablo. A final editorial board consisting of Mr.Hailstone, Mr.Davey and myself arranged the individual paragraphs into a logical order, after they had been checked by the section leaders. The journal is divided into sections by numbered underlined headings. Within each section, names of individuals generally precede their contribution. Some members preferred to submit pages of ready typed copy for their contributions.
Some of the rather detailed lists of internal papers produced have been edited.
The last year has been exhausting. The pitiful staff levels have exhausted the group. The addition of Julian Gallop and Jacky Hutchinson has been invaluable but the loss of Graham Robinson a great pity, just as he was becoming very familiar with the system. The work is now vulnerable with only one system expert - Dave Toll - as Paul Bryant now spends most time in managing the many machines. It seems a pity the project cannot be reasonably staffed for one reason or another. It is also a pity there is so little effort on developing applications software in the Laboratory. It is appreciated that the heavy demands for system development time reduce the machine's availability. However, with the network, it is now possible to use anyone of 6 machines which are currently underused.
Progress has been remarkably good. Machines have been very successfully installed. In addition, the users seem very pleased with their gifts.
The hardware on all the sites has been good. As we have learned to expect, most machines have a failure of some sort shortly after delivery. However, the failures follow no pattern and the machines are proving very robust. The CPUs themselves have been highly reliable. Of particular note is the fact that no files have been lost due to disc failures. It has, on one occasion, been necessary to return to a backup of the system disc at a remote site.
The Benson Plotters have proved a good buy. None of the plotters has yet failed and they continue to give good accuracy. In the agreement with Benson, certain changes to the interfaces were agreed. These changes were to check sum records and include double buffering. This is now complete. However this has little effect on the use of the device but has improved its performance. The interface specification is now stable. The plotters in the field have not been upgraded yet, although this will be undertaken shortly.
The sites have shown little demand for a larger range of peripherals beyond data collection facilities which cannot be provided except by defining and providing a 'standard interface' such as the FTP-B on an asynchronous interface. Certainly the decision not to provide card readers and magnetic tape seems correct except in the case of Cardiff who have a good case for magnetic tape.
The speeds of the multiplexer have now changed from 75/110/300/1200 bits per see to 300/1200/2400/4800 bits per sec. 110 now seems obsolete and the 75/1200 speed is provided by a retiming system. It was found almost impossible to automatically speed select over the range 75-4800 bits/see due to hardware limitations, but 300 to 4800 is satisfactory.
About 12 months ago a programme of software development was produced. All items are now complete.
It was in retrospect a mistake to mount Version 2.3 + batch of OS4000 in late 1978. It proved unreliable and GEC showed little enthusiasm to support it. The development had advanced considerably before this was fully appreciated and the project was continued. The dissatisfaction culminated with a difficult high level meeting with GEC which improved matters but unfortunately rather after the event! The system was made reliable round about Christmas and since then has been highly satisfactory. It should be pointed out that the unreliability did not impact the other sites noticeably. It must not be forgotten that the decision to move to 2.3 + batch has given sites added facilities such as batch, as well as avoiding a painful files tore conversion at a future date.
The most exciting development has been the X25 network code. This was done in co-operation with C&A Division. The network X25 node has been generously provided by C&A Division on a GEC 4080. The code was undertaken by Andrew Dunn who also provided the X25 call and link level code under OS4000. He also provided HASP under X25. The interface to the files tore (now called MAFS) was provided by Dave Toll. This was a very extensive system and it includes a network spooler and user information systems to ensure the system is usable. File Transfer Protocol was a joint effort started by Dave Toll and finished by Graham Robinson. Interactive Terminal Protocol into OS4000 was written by Paul Bryant, and in the outward direction was contributed to by several people, in particular Tony Salter. Even with the broad authorship, the result is a superb well-integrated product of which the Laboratory can be proud. It is now being brought into use and has transformed the support that can be provided to a site from Rutherford and will contribute to the feeling that the machines form a community with a community spirit.
The wide ranging programme of system development made it necessary to keep a master copy of the operating system, so that equivalent and reproducible systems could be produced for all remote sites. whenever a phase of system changes is complete, a list of altered files (provided by system developers) is used to make updates to the master copy and a comparison procedure indicates whether there are any omissions in the list. A largely automatic system for performing the update was produced early in the year and the masters thereby produced have been used for new operating systems at all the remote sites.
There have been many minor developments in the operating system. It is a pleasure to record our gratitude to Phil Gladstone who produced the Benson Plotter driver code which is a fine product with very good facilities. The plotter is driven as a spooled device which allows plots to be produced when the plotter is switched off. Good heading facilities have been provided as well as pen changing facilities, which are virtually foolproof. Phil was also responsible for the error trapping code which has been used by the FORTRAN system to give error trace backs and re-entry facilities under all circumstances. Tsang - a student - wrote the FORTRAN modifications. Most of the other changes have been to make minor improvements and to remove bugs.
There have been minor changes to several utilities and a few new utilities of a relatively minor nature. In fact the set of utilities seems quite good although some rationalisation would be useful, but GEC should do this.
The GEC at Rutherford used to be connected to EPSS. This connection has been removed as the hardware was required for the X25 connections, also little use was being made of it. One could say it had served its purpose well and provided valuable experience. .
This is provided by Jacky Hutchinson (who we have been fortunate to recruit) and Julian Gallop.
The principle of supporting these machines from Rutherford has proved successful. Sites have been most appreciative of this type of support and surprisingly uninterested in tinkering with the system or developing it themselves. In the early days the lack of communication lines made support difficult but just about possible. With lines, the situation is transformed and receiving queries and posting replies is technically easy. Every day each machine is logged into and the 'Support file' printed and action is taken. Of course the 'gripes' are then filed. In fact the requests for support have been at a modest level and have been reasonably straightforward. Few software errors have been discovered by users. It is remarkable that we now have 6 machines supported by 11 people (this includes all operators on sites and Rowland).
Two manuals have been produced. The Primer (issued in November 1979) describes facilities useful to a new user of OS4000 as well as outlining a few problem areas likely to be encountered early on. The Reference Manual describes the system in detail and was issued in a provisional edition in May 1979.
It is the intention to supply software via the network. To date most software has been delivered 'by hand' or rather disc, due to the network being deficient in lines and modems. Also the full network system is only just available. However, preliminary tests using HASP at Bristol have proved the feasibility.
Username insertion is a joint exercise. ICF put up SRC users as it is easier to do it via the network. The sites insert the usernames for their 20% of the machine. Safeguards in the code prevent the site tampering with SRC users and in fact they are not able to put in SRC usernames. University provided usernames must have a U on the third character.
The machines at sites were delivered as follows
Installations have been amazingly successful, but not easy. The technique of installation and indeed the preparations for installation have been refined during the period and procedures are now exceedingly smooth. Nonetheless, each site has had its peculiar problems and has to be treated with an individual touch. For example in some cases the room is horrible and in another, politics are sensitive.
In all cases actual installation has taken 3 or 4 working days, although in some cases weekend and Bank Holidays have increased the real time. In addition the physical lack of staff has been a problem. Once or twice it has been a close shave between a room being prepared and delivery. In fact, delivery was postponed for a week in one case. Room preparations have in general taken twice as long as predicted by the site. Six months seems par for the course.
Actual installation has been carefully supervised. Rowland normally supervises actual installation on the first day when he also lays cables and checks terminals. The second day is taken up with acceptance tests and more cabling and testing. On the third day the software is mounted by Jacky or Julian while Paul makes a good will visit to tidy up the politics. On the fourth day a service is provided.
The close spacing of installations has placed severe strain on manpower and it is appropriate to pay tribute to the hard work and many hours of overtime spent on the project. Ideally it is not reasonable to expect to install more than one machine every 2 months with current staff levels.
Training has been surprising. Many users have managed reasonably unaided. Paul Bryant normally gives an introductory talk which is not well supported. Attendance has varied between 6 and 50 with a mean of 20 or so. There has been no demand for advanced courses or repeat talks. This is not necessarily because Paul is a lousy lecturer but is because most of the users are already well acquainted with computers and assume that they can start work more or less unaided. This has proved correct as users have found the machine easy to use.
Use of the machine has built up slowly. Even now, Bristol is only 25% used. It is very unusual to be given a machine which needs little or no running up by a team of programmers and thus has relatively unrestrictive growth. It seems that users need time to 'believe' that a machine has arrived, longer to believe it works reliably, and then longer to make the effort. Nonetheless, use of the machines is generally increasing.
So far the machines seem very adequate to the task demanded of them. In many cases FORT2 has proved adequate and the worries over large programs have not (or at least not yet) materialised. The most significant demand apart from FORTRAN and GINO has been for BASIC. There is a significant demand for other languages and systems but each site has a different request. Most demands can be met.
A grave problem until recently was the lack of a reference manual which was starting to raise questions on the site. In retrospect it was not a good idea to divorce manual production from the overall project management.
It is a pleasure to record the valuable assistance given by Jim Nisbett of GEC before and during installation. It was a wise plan to have the same installation engineer in every case.
The operations staff are responsible for the general day to day operation of 4 on-site machines, 2 PRIME400s, GEC4070 and the DN82, for checking out ICF terminals before they are dispatched to the end user. They also assist in the installation and initial operation of MUM's in the field.
During the second half of 1978 both of the operations staff left. When the new staff joined there was little or no overlap and they were thrown in at the deep end; this was a particular problem as the program of installing machines in the field was just taking off. They managed, to their credit, without any disasters.
Most housekeeping jobs are done via BATCH jobs on the PRIMEs, or at the beginning of the day shift. Some overtime is necessary to take essential work out of the prime shift, this amounts to some 3 hours a week. Major files tore reshuffles are done over weekends periods, but these are rare events. The reconfiguration of the PRIMEs meant that the backing-up procedures had to be revised; most files are backed up at leased twice a week. It is hoped that the arrival of the PNCs will help backing up across the link.
Some 80 terminals were checked out during the first 3 months of this year. This peaks at the end of the financial year; it also clashes badly with the installation of the MUM's. The lack of a permanent home to receive, check and repack terminals has meant many wasted hours moving equipment around. The program to checkout terminals is well worth while, as many as 20% require some tubes, blown fuses, interface cables to further attention. The faults range from broken loose wires, faulty boards etc. Some terminals require be made, others require power plugs to be changed.
Two major software efforts now remain. Firstly, version 3.0 of OS4000 is with us but there is no hurry to mount it. This should be mounted by October. It will give few user benefits but will make the site more 'standard' as regards GEC software. The second project is to produce an archival system. This should be finished before Christmas.
Minor projects are:
The Primer needs drastic revision and this is scheduled to be complete in September. The Reference Manual will have similar treatment, scheduled for completion by the end of the year. Revisions are needed due to user comment and system changes.
The next 6 months will also be a period of consolidation now we have a very brief breathing space.
There has been a demand for connecting the computers to data collection equipment and other devices. After much thought it seems the best method of connecting devices is using the asynchronous file transfer system currently working between the PRIME and GEC. PDP11 and LSI/11 versions existing currently. An Alpha 16 version is under development at Cambridge. It is hoped to produce Z80, 8080 and 6800 versions and to this end a Horizon Z80 system is being purchased.
An IEEE V24 interface has been purchased to examine the problem of connecting IEEE devices to the GEC computers.
A paper tape reader/punch on a V24 interface has been purchased to examine it as a substitution for the GEC version.
After many years, the DoI Advanced Computer Techniques Project has been disbanded. This organisation supported many advanced projects such as the DAP and the content addressed file store as well as many other hardware and software products. Paul Bryant sat on the ACTP software committee and currently looks after the DAP and Toltech (Tripos operating system) projects and Dave Toll looks after the Demos project. These projects will continue until their natural termination.
The DECNET 2050 workstation product has been a bitter disappointment. This product is being produced at Hatfield Polytechnic under an EMR agreement. 18 months ago the project looked promising and was demonstrated. Since then, it has been in a continual state of 'two months to completion'. Initially it was unreliable and lacked some facilities and there has been a very slow improvement. Currently it has been discovered that it has throughput problems due to it being on a complex network rather than the simple one at Hatfield. A further problem is that the DECNET network has reached its maximum size. Any increase will require more store on all the DN80s. One breathes a sigh of relief that the SRC network is not based on DECNET.
Early this year 6 TAC terminal switching systems were ordered from Imperial College for location on sites. These should be delivered soon but the world wide component famine seems to be delaying them.
A small amount of time has been spent by Dave and Paul in investigating the NSC high speed bus as a possible means of separating computing from storage. A visit was made to discuss it with John Downs and Bill Williams at ULCC.
TALKS GIVEN
REPORTS
Report on CRAY / Seminar: 6 June, J Gallop
The multithreading File Transfer Protocol process is now completed and can communicate with the Prime400 0 or the SRC X25 network. Testing with remote GEC's will be commencing shortly. The implementation is to the full ACD 'Green Book' specification and in addition includes the recommendations of 2 Apr from the Data Communications Protocol Unit.
Much time has been spent on a computer tender and assessment on behalf of a Panel of the ASR Board. It is proposed that six powerful mini-computers are purchased, with colour raster displays, for astronomical image processing. Five will be spread around the country and connected in a star network to the one at Rutherford. The laboratory will be responsible for installation, software co-ordination, etc.
For the computer systems, an Operational Requirement was issued by Appleton Laboratory. We were responsible for an assessment of replies to this which led to a shortlist of PRIME, DEC, IBM and ICL. A tender document was issued from here and PRIME, DEC and ICL replied to tender. A report on the tender replies was produced in June and a Council decision on the machine to purchase should be made on 18 July.
The tender for image display systems was also issued and co-ordinated from here, although assessment is being done at RGO, with the help of Alan Francis (ex Atlas) as a consultant.
GWR has taken over Systems from LOF, who is clearing things up prior to States trip. Setting up the dual service here has proved troublesome and still not yet finished. However the Primes are linked to each other and to the ICF Prime at Nottingham using PRIMENET. No decisions are yet taken about how or when to link to SRC net.
User documentation is in a woeful state and there is little effort to do much about it. The Nottingham HELP system is being introduced, and it is hoped that information will go into this in parallel with new manual production.
An incremental dumper system is being implemented by two students from Brunei (Chris Kennedy, Ralph Mitchell). The dumper part is finished and work is now proceeding on the restore. This should make the PRIME filestore back-up far more efficient.
A command to print RUNOFF files on various devices, including the A4 Diablo, has been implemented by a student (K Tsang). He is now working on amendments to the line printer spooler.
As secretary of this, I was responsible for the agenda of its annual conference (in Seillac, France, in April). It thus had three RL speakers (Phil Newton, Anton Walter, Roland Rosner). A write-up is in a 'PRIME NOTE'.
There has been one meeting in the period. Cynthia does all the administration and the lack of a Swindon office secretary is probably now an advantage.
Responsibility for this system was assumed on 15Jun and efforts are currently being directed to curing various problems with HASP and POST resulting from the linked machine configuration.
The slow arrival of new versions from Hatfield Polytechnic has resulted in a slow but working system. The next version due shortly should speed things up sufficient to let it loose on users to find out how reliable it is.
A second PR1ME 400 computer was delivered in March to increase the number of interactive users supported simultaneously. This second machine was partly paid for by technology division to provide an interactive service for EBL users. During this quarter the machines have been connected together by a slow speed link as the high speed link hardware has not been delivered yet. PR1ME say that this hardware is being tested in this country now. Work has been done on enabling the two systems to act as a dual processor system.
FTP has been rewritten to include a call level between the asynchronous protocol defined by Dave Toll and FTP version B. Additional facilities have also been included, such as the ability to send plotter files to the GEC Benson plotter.
The Benson plotter has been upgraded to be compatible with the GEC plotter. This has entailed moving it from the sac to a normal AMLC line and rewriting the despooler.
The code from the EBL PDP 11/34 for FTP has been mounted on a LSI-11 operating under the rt-11 operating system. This process involved writing a mini-FTP to boot the main FTP and a TTY emulator.
During the last three months I have mounted a new version of PRIME software REV16 complete with various modifications. Also the introduction of a second PRIME 400 has caused me quite a lot of work keeping the two machines' system directories in a similar state. In addition I have also made several minor modifications to PRIMOS.
The performance monitor is now almost fully automatics and requires little operator intervention. I have also been preparing documentation describing how it all works.
PR1ME Users are documents issued to each accredited user of the PR1ME 400 and they describe features and software packages available on our machine and which are formally supported. PR1ME User Notes, on the other hand, are only distributed internally. These tend to be more technical and are issued by the person concerned to describe recent mods, new features etc on the PR1ME. PR1ME Users usually begin life as PR1ME User Notes and are produced where appropriate when the feature in the PUN is reasonably safe for external users to begin using it. PR1ME Technical Notes also appear from time to time and as the name suggests they deal with the more technical aspects of the PR1ME such as new hardware and usage accounting and budgeting.
The graphics facilities on the PR1ME are expanding steadily and are described in a recently issued set of Graphics Users. Where a feature is described in both a PRIME User and a Graphics User the latter should be regarded as the more up-to-date description.
When not churning out Documentation, putting new users on the system, or answering queries from people who sometimes haven't bothered to read the manual I am on a Brunel course, working on a Database project, mounting new software or generally keeping the users happy.
During April the mailing list was updated. A large number of changes were made to the computer files (mainly removing duplicates, replacement of out-of-date addresses and inserting new readers. The latter included new members of SRC's Engineering Board and it committees, subcommittees and panels. The mailing list is now in a good state and could be the basis for a rationalised distribution system for other publications, in the ICF and perhaps elsewhere. About 1800 copies are now printed of which about 1650 go outside Rutherford to about 1100 addresses.
In May the 9th issue was prepared, printed and dispatched. Since then, proposals have been put to PMICS about how material should be identified for inclusion in future - PMICS will be involved more than in the past.
I am the secretary of SIG CAAD, which met in Bristol on 24May. It approved actions along the lines agreed at the last open meeting; these include:
Since then I have been implementing these decisions.
I have attended meetings of a committee on 'Auto-generation of programs for construction site management', chaired by Professor Trimble of Loughborough University. This Committee was set up under an SRC grant. I have now dropped out owing to pressure of other commitments but one of the other SIG members is to sit on one of the sub-committees.
Work is proceeding under four main headings:
The section is still one under complement and this is delaying progress.
The evaluation exercise on the four FE packages CASAS, BERSAFE, NONSAP and SAP IV) was completed in this quarter. This has been a major undertaking, requiring not only mounting of the packages, but an extensive programme of test problem generation and reconciliation of the results obtained. At a recent meeting of the FE Steering Group, it was decided that the ASAS and BERSAFE systems should be supported.
User-oriented macros to access the 360/195s from the PRIME and GEC machines are being written and will be released shortly. Real users are already on the system and are requiring support, particularly Birmingham whose current runs are taking about one hour of 360/195 cpu time and a correspondingly large amount of disc space!
The Library is at two levels - Level 0 (basic building blocks) and Level 1 (example programs). The initial release will be ready for the FE User Meeting on Tuesday 18 September.
The code for the initial release of the Level 0 Library is complete, the major programming work in the the quarter being the addition of an error-trapping mechanism, and some extra routines needed for Release 1. Documentation is two thirds complete and has been the greatest consumer of effort. A program has been written to post-process output from POLISH (supplied by D A Duce) , and another to process RUNOFF output files so that the FR80 can be used to prepare documentation. The SYMBOL routines of K M Crennell have been extensively used for this part of the project.
At Level 1, we are not so well-advanced, principally due to the difficulty of producing an acceptable format for documentation in a complicated area. One of the example programs has been used as a test bed to define a format for the Level 1 documents. This format has now been accepted, after some iterations, by the FE Steering Group. Code for all the programs in the initial release is complete and work is progressing on the documentation.
Most of the work here has been involved with the Working Party on Pre and Post Processors, usually referred to as P4 for obvious reasons. P4 has met several times since June 1978 and at the last meeting a paper intended as a prototype for a future ICF paper requesting effort and money was discussed. Comments on the paper by other ACD staff are gratefully acknowledged.
As a parallel to this activity, the package-independent pre-processor FEMGEN has been obtained for evaluation purposes and is likely to be adopted as a standard ICF pre-processor once the contractual problems are resolved.
KR is the contribution to the C06 (Summation of Series and Fourier Transforms) chapter of the NAG Library. A large body of software to perform the FFT has been obtained; this offers four separate methods, allowing the user to trade-off cpu time against storage. Considerable work has been necessary to bring the software to NAG standards.
The major effort this quarter has been to run comparative timing tests for the routines. This has been done initially on the PRIME 400 and is now being run on the GEC 4070. With NAG Central Office's help, some runs will also be performed on a CDC 7600. By using different machines, we hope to be able to give accurate guidance to NAG Library users on the choice of subroutine. Documentation is also (inevitably) proving a great time-consumer. The GEROFF system on the GEC 4070 is being used.
23 May, K Robinson "Survey of Finite Element Software - A Consumer's Guide to Availability, Costs and Scope" at Polymodel 2, Teesside Polytechnic
The April round of SRC grants was the heaviest the section has ever seen. Over 100 full grant applications were sent to ICF for comment. A high proportion were actually asking for use of the ICF as well as many which wanted to have their own computing equipment. A reasonable number wanted the now common single user minicomputer for data acquisition of real time control of experiments and then an allocation on the ICF to analyse the data or model a system.
The process of assessing these applications and the relevance of the ICF to their computing requirements has consumed the majority of JB and BGL's time and still the job is not over.
The task of providing the RL Management function for the DEC10s and the external ICF minicomputers is carried out partly by this section. The major event towards the end of the period was the budget discussions with UMIST and Edinburgh installations, the net value of which is reaching nearly £300K.
General responsibility for ICF terminal pool (despatch retrieval and problems). This currently stands at 333 devices (276 of which are scattered around the Universities, 46 of which are on the Rutherford site and 11 as yet unallocated) . 233 of the devices are maintained by Cable & Wireless. Monthly amendments containing details of locations and contacts have to be sent to AERE contracts, who oversee the contract.
Some ICF users are allocated funds for dialling into host machines prior to installation of leased lines. They produce periodic telephone accounts which have to be checked against machine time used before being paid. In the current quarter only 4 bills have been received amounting to $685.06
Moving users files around the ICF network. Generally from their current machine to a Dec-10 via the IBM360 (when it lets me!!) The 2 main users of the facility have been Cranfield and Oxford. The main problems produced have been because a standard utility does not exist on all ICF machines and the unsatisfactory nature of the network for moving large numbers of files.
(J.E.HAILSTONE, P.D.ATHAWES, P.KENT) A detailed report has been prepared showing the frequency distributions for the common variables, elapsed time, cpu etc. based on details of each login recorded. Preliminary analysis shows that the demand for CPU time is complex, with important implications for response time and total capacity to be expected from the machine. Some theoretical work is leading to the prospect that, from the normal monthly accounting figures, some critical parameters can be derived which will help in monitoring the quality of the interactive service.
Mr A Payne completed his study of the user profile at Edinburgh and UMIST and a report has been submitted.
(J.E.HAILSTONE) After the selection panel at St.Catherines' had met a candidate was selected and an offer is being made for the post. No suitable candidate was found for the St.Hilda's fellowship and this is to be re-advertised later this year. Trinity Oxford have asked to reopen discussions about a possible fellowship with the college.
Initial work suggests that a small number of users make demands which cannot be satisfied by the present PRIME computer without degrading the service for the rest of the users. The extent of this degradation of service is being examined. It may be that the type of work being done by these users really requires a more powerful computer.
Statistical models are being constructed of the behaviour of users of the PRIME. It is intended that the work should be extended to analysis of the DEC10 machines.
The Artificial Intelligence community fit rather unhappily in the ICF. The existing resources seem inadequate to their needs while anything that the ICF is able to offer is likely from their point of view only to make matters worse.
SPSS and ALICE are now available on the PRIME computers. CENSTAT should be available shortly and a PRIME version of GLIM is being prepared by Loughborough University.
R E Thomas, D C Sutcliffe, H F K Yeung, M R Geary A D Bryden, J Jezek, P A Dewar, I D Benest, R K L Gay, J A Mclean
The section has three main groups: graphics, AP120-B evaluation and Circuit Design. In addition, there is an interest maintained in ICL's Distributed Array Processor, and some involvement with Starlink. During the quarter, Dr Robert Gay joined the Circuit Design (DACD) group but there are still two staff posts to fill there.
This quarter has seen a consolidation of the graphics software on the two machines. This is reflected in the birth of two new series of external user notes, PRIME GRAPHICS USERS and GEC GRAPHICS USERS, which replace and supplement the graphical notes in the USER series of each machine.
GINO-F handlers have been provided on the Prime for the Benson plotter, Imlac Dynagraphics, and new colour GOC, although the latter requires further development to make more use of the facilities available. The Benson and GOC handlers have been provided on the GEC but the Dynagraphics handler awaits the acquisition of the GEC source of GINO-F since it is a modification of the PDS4 handler. An EMR agreement with Leicester University is proceeding satisfactorily and will provide a 'full refresh' GINO-F handler for the GOC. This quarter has seen a consolidation of the graphics software on the two machines. This is reflected in the birth of two new series of external user notes, PRIME GRAPHICS USERS and GEC GRAPHICS USERS, which replace and supplement the graphical notes in the USER series of each machine.
Work is progressing on the first phase of the provision of the high level graphics library, FNGRAF, to work with FINGS. 3-D histogramming and area shading will shortly be available on both machines. This quarter has seen a consolidation of the graphics software on the two machines. This is reflected in the birth of two new series of external user notes, PRIME GRAPHICS USERS and GEC GRAPHICS USERS, which replace and supplement the graphical notes in the USER series of each machine.
Some work has been done in FINGS, mainly the provision of a handler for the Benson plotter, although this is not yet publicly available. This quarter has seen a consolidation of the graphics software on the two machines. This is reflected in the birth of two new series of external user notes, PRIME GRAPHICS USERS and GEC GRAPHICS USERS, which replace and supplement the graphical notes in the USER series of each machine.
With the increase in the number of MUMS using RL software, there has been an increase in the amount of graphics user support necessary. This quarter has seen a consolidation of the graphics software on the two machines. This is reflected in the birth of two new series of external user notes, PRIME GRAPHICS USERS and GEC GRAPHICS USERS, which replace and supplement the graphical notes in the USER series of each machine.
New demo programs have been provided on the P400 for the IMLAC and GOC's. A new PROM is still awaited for the IMLAC.
Contact is being maintained with Manchester Graphics Unit, to see whether the raster display they purchase, and the software they write, is of use to ICF.
Seventeen of the twenty six T5600 GOCs have now been sent to users and another four are in use at RL. There have been a few initial problems, including their use with model 20 modems and the size of characters. However, the former is now solved and the latter is being solved by the incorporation of a scrolling alpha store with larger characters.
The three new colour GOCs have been delivered but little work has been done on examining the polygon fill capability.
It has been demonstrated that the Tektronix 4632 video hardcopy unit can be used for taking hardcopy from both colour (using grey scales) and monochrome GOCs, instituting the copy from the GOC keyboard. Some of these will probably be purchased this year.
There have been demonstrations of other graphic devices including the Tektronix 4027 colour terminal and the Retrographics graphics add-on board to the Lear Siegler ADM3A alphanumeric terminal. The former confirmed our opinion that the colour GOC is superior and the latter created some interest as a cheap graphics terminal. Note is being taken of Sigma's ARGS (Advanced Raster Graphics System) which is under development.
ICF is represented on the NAG Graphics Working Party, responsible for designing a graphical interface to any graphics package and for providing high level graphical routines, which use the interface, for inclusion in the graphics chapter, J06.
As a prelude to bringing DRIVER, the new control software for the FR80, into operation, a series of releases has been agreed with FR80 Operations Group. The first release, handling 35mm black and white film only, was introduced in June, and has proved very successful, being faster than the previous software. Test runs on the 16mm black and white version are also being conducted in parallel to the normal workload. Subsequent releases will include new character facilities, font changing, colour and fiche. Finally, work will be done to fine-tune the software.
The introduction of the new character font facilities has an effect on existing software on the 195, and so it is necessary to proceed with full consultation to avoid any difficulties. In particular, the new font changing will need host software support. Also, to incorporate all the proposed features it will be necessary to increase the fR80 store size. The case for this is being prepared by fR80 Operations.
The investigation is continuing into the usefulness of the Floating Point Systems AP120-B Array Processor to the ICF. In April a new version of the FPS software which contained the long-awaited AP Fortran compiler and a new loader was installed on Prime 400A. (We are the first site in Europe to receive this software).
ADB wrote a simple macro to use AP Fortran together with some preliminary documentation and circulated these with the relevant FPS documentation to major users of the AP120-B. The new software has been discussed with internal RL staff and its use encouraged by communications with users in Reading and Sheffield, and through visits to Oxford Bristol and Cambridge.
It soon became apparent that there are 3 major problems in the FPS software (plus several minor ones). First, the compiler and loader run extremely slowly; secondly it is extremely time-consuming to load the program to be executed into the AP from the Prime and thirdly there are some bugs in the AP Fortran compiler some of which cause the AP120-B to go into a loop.
ADB and JJ are investigating these problems and are trying to improve the efficiency of the FPS software in several different areas. A meeting is being arranged with FPS to discuss these and other problems.
During the quarter ADB has attended 2 courses in Mathematical Methods in Computing at Harwell and SRC Promotion Interviews at Central Office, which involved an absence from RL of about three weeks.
I learned how to work with the Floating Point Systems AP120-B Array Processor software. I also attended a one week programming course organized by FPS in Bracknell.
I wrote several tens of subroutines for elementary matrix manipulation. The idea was to augment the FPS Maths Subroutine Library for it to be suitable for implementing finite elements algorithms. The implementation of existing mathematical packages in the AP120-B is not a straightforward process of recoding, some sort of algorithm restructuring is necessary to utilize the array processor efficiently.
I have temporarily abandoned the work in numerical algorithms and started the more urgent work: to try to improve the efficiency of the basic software. It becomes apparent now that dramatic savings of time consumption can be reached by changing of some file formats.
(P.A.Dewar,I.D.Benest,R.K.L.Gay,J.A.Mclean)
The process of turning the SRC-approved proposals for ECD on the ICF into a system that engineering researchers will be able to use is now well under way. From the many hundreds of ECD computer programs, about a dozen have been selected as worthy of incorporation into a design system. Useful contacts have been made in Europe, the UK and the USA, and considerable interest is being shown in our project. In the field of Digital Design, the combination of the ICF, the EBMF and the FR80 is unique and the envy of most researchers in the field.
As well as the work listed below, the following major problems have been addressed:
The DACD disk area (ECD) was established and is now maintained by I.D.Benest.
This has been mounted on the UMIST DEC-10.
An IMLAC PDS-1D graphics monitor was substantially changed to run in the IMLAC DYNAGRAPHICS.
I have investigated the feasibility of developing a Digital Design System based on a machine readable library of digital integrated circuits and MOS substructures. Sadly, I discovered that this, the first stage in a comprehensive Digital Design System, required manpower and software which is unavailable to the DACD group at present.
In preparation for the long term Digital Design project, I have carried out a brief preliminary survey of what university researchers require in a digital component library and have run some test software on a PDPl1/34 running UNIX at UCL.
Within the next two months the DACD group will be able to release the following digital packages:
The DACD disk area (DACD) was established and is now maintained by I.D.Benest. The following tasks have been completed:
The DACD on-line HELP system was established to improve the environment in which the DACD user works and also to reduce the time necessarily spent on user support.
SPICE2: A general-purpose circuit simulation program, recently mounted on the IBM195, was tested. Analyses of typical circuits of various sizes were carried out. Amongst circuits analysed were:
In a 1975/76 survey of Computer-Aided-Design and Analysis programs carried out by a team from the University of Florida, it was reported that erroneous results were obtained using a version of SPICE2 on two of the test circuits concocted by the team. The same circuits were analysed using DACD's version of SPICE2. No errors were apparent and the results were correct.
Comparison of accuracy and speed with ASTAP was/is being attempted. There are difficulties because of the different transistor models used in ASTAP and SPICE. Conversion of some parameters is possible, but the availability of other parameters depend on the types of measurements manufacturers have used to characterise their transistors. For certain circuits such as the NAND gate, SPICE appears to be much faster. ASTAP required 86secs of CPU time to do only 130nsecs of transient analysis.
Currently conversion of NAP2 to run on the PRIME400 is being attempted. Runs on the IBM195 indicate that NAP2 is extremely fast for small circuits. Successful conversion would be a step towards an interactive analysis capability on the PRIME400.
Monitor level access to the following IBM360 circuit analysis programs: ANP, ASTAP, NAP and SPICE was introduced. This included monitoring facilities to gauge the usage of each program. Considerable time was spent on providing user support for ASTAP.
There is a continued involvement with the ICL DAP FORTRAN user group, where the current state of the language is discussed, suggestions for improvements made, and plans for the future formulated. Currently, the first production version of the compiler is on schedule, but the hardware has been delayed, and is now unlikely to be delivered to Queen Mary College, London until December.
There has been some slight involvement with the Starlink project so far, but this might well increase in the future. Work was done in getting the FORTRAN benchmarks to run on the P400. In addition, the tender exercise in raster displays and the subsequent software written is likely to be of great interest to ICF.
(M.F.Chiu,M.E.Claringbold,T.W.G.Dunwoody,R.J.Waters)
Two years' work converting the SRC Grants System, from Batch on the RL ICL 1906A to Interactive working on the 2904, was completed this quarter. The system went live in the last week of June 1979. Work in the last 3 months has concentrated on producing full documentation, implementation of Data Base checking utilities and statistics-gathering programs, regularising the mechanism of archiving, and performance monitoring. The large task of converting the serial files holding grants information to a structured TOTAL Data Base was achieved with few eventual problems - although it required more than 22 hours of elapsed machine time for the two final conversion runs.
The 2904 computer now has 90K words of user store and the central hardware appears to be very reliable. The operating system is now EXEC 1S/MTS and although it contains some ICL idiosyncrasies it also contains new features (over EXEC 1/ETS2) that considerably facilitate both operating and program development.
With the completion of the main Grants programming (and some 50 - 60 attendant Report programs), requests have already been received from the users for 'enhancements'. Two of these are currently being implemented: Automatic Indexation of staff salaries to allow for the cost-of-living increases, and the Engineering Board Subject Classification scheme to allow interactive retrieval of grants by keywords.
Preliminary work has just commenced on the Studentship Award System, also to be re-implemented on the 2904. Although there is still considerable uncertainty about a detailed specification for Awards, the groundwork of Data Analysis for the Awards Data Base can be carried out as a profitable exercise.
A number of the facilities of PRIME DBMS have now been used but there are quite a few still to be investigated (particularly privacy controls). There are 2 main databases under DBMS and a number of small test ones.
The computing divisions telephone directory database now contains the data for the 2 divisions and has a suite of programs associated with it. These allow input of data, updating, retrieval and list production. It is at present updated sporadically.
Rev 16.2 of DBMS is now on the PRIME, it cures a few bugs and has improved performance. Timing tests with the telephone directory database show a saving over the Rev 15 version of about 28 in I/O.
Mattie Keane and Ann Walter have been setting up a database with names and addresses of PRIME users for documentation distribution. This exercise was delayed by the discovery of undocumented restrictions in PRIME DBMS, but is now at the stage of testing application programs.
There is general interest both amongst ICF users and non-ICF people in databases, data base management systems and data structuring. The difficulty at present is that nobody knows quite what they want and it is not easy to decide what type of software is best in any particular case. A proper evaluation in each case will require a lot of time. A pilot study has been started with Cyril Balderson on the possible uses of database software for telecommunications information.
Some small programs have been written to test MIDAS and find out how to use it. These can be made available as example programs together with the introductory documentation to users starting to use MIDAS.
A small amount of work has been done trying out the facilities of STATUS.
Most of the time has been spent on 3 data bases, one of which can be used for demonstration purposes. There has also been some investigations into the general facilities of STATUS, and also into the sizes of the data base files that are generated by STATUS.
All of the interface and utility routines are now complete. Work is continuing on the Editor program, which is to be used for writing, deleting and updating records, but has recently come to a halt with an 'OC1' error, which is being investigated by C&A. Work on the documentation, using Electric text layout facilities, is continuing.
Hard copy for the main text of Vol 10 of the series 'Molecular Structures and Dimensions' was sent to the printer by their deadline of 23 April,1979. There were the usual last minute panics, caused this time both by the user and FR80 Operations. The user wanted a new input tape format, and automatic contents page generation. The FR80 had a new tube installed in March, with different operating characteristics, thus causing changes in the intensity settings of PURL. 371 pages of hard copy were made, using 45minutes of FR80 time, and 2minutes of IBM360/195 time for the production run.
In preparation for mixing crystal structures with text some day the program PLUT078 was mounted on the IBM360/195 and structures drawn on both microfiche and hardcopy. I think the final program will need to be interactive so I tried to implement it on the PRIME. Unfortunately, the FORTRAN is not as portable as it should be, and so far the main user has been Len Ford, who was looking for a program on which to exercise the new Debug package.
The subroutine SYMBOL already mounted on the PRIME400 and IBM360 has been speeded up by using locally available logical and shifting functions. It has also been mounted on the GEC4070. The same text files run on all 3 machines, should now produce the same output drawings. Preliminary documentation has been distributed to User Support.
The interactive (reactive?) program intended to produce FR80 output, and simplify the production of the front covers of RL reports, was working on the GEC4070 at the beginning of the quarter using FINGS.
Since the FR80 driver for the GEC4070 was not available, we tried to convert the program to use GINO-F on the IBM360. This has been a useful exercise for the student, but has so far not resulted in a production program for Photography.
ADDRESS LABELS PRODUCTION
The address labels program using FAMULUS on the IBM360, has been made available from the data editing GEC4080. Multiple copies of labels are now possible. A further procedure LBLSINDX has been set up, to provide statistical summaries, and simplify the problem of maintaining accurate addresses.
PHOTOGRAPHIC CATALOGUE
Following discussions with H.Edwards(Photography), a procedure was set up in the data editing GEC4080 to make indexes in several orders using FAMULUS. V.Walker has been learning how to use the 4080 by putting the contents of the Atlas Slide library into the file. This small sample will be used to test the logic of the procedure, before starting on the main library for Photography.
Although we are no longer supporting COCOA, I still receive requests for information and advice. This quarter I have answered letters from University staff in Australia, Spain, Holland, Cardiff and Stirling. A copy of COCOA was sent to the University of Iceland. I also spent a morning discussing computer techniques for analysing textual material, with research staff from the Department of Education at Reading University.
I attended the ICL launch of their Word Processing system held in April. I also attended a 3 day programming course on the Delta Data 7000 series intelligent terminal development system. The system delivered here in march was defective, an almost complete replacement arrived in late June, but we are still waiting for some of the software. It contains a programmable microprocessor, 2 floppy discs, a printer and an alternate character generator in RAM. We have ordered word processing and font defining software, and hope to use them to study the problems of using this type of device for scientific text, when connected to our mainframes.
Major activities in the last three months were the preparations for the May DCS Panel and the June Computing Science Committee meeting. The DCS Panel on 3 May 79 left several actions on the Academic Coordinator with respect to clarifying certain points about some grant applications before they were considered by CSC. These actions were completed on schedule.
The other major activities during this period were concerned with the Equipment Pool, RL financing of DCS and preparation for the 3 July 79 DCSP meeting.
The questions raised by the DCS Panel about the grant applications of Bull & Sayers, Welsh, Osmon and Bennett were discussed with the applicants and their answers put before the CSC on 4 June 79.
The Data Base applicants (Deen, Stocker, Shave & Flower) have all been contacted and they all think that a meeting prior to revised submissions is a good idea. It has not yet been possible (22.6.79) to fix the date for the actual meeting, in spite of lengthy negotiations.
The distribution of the existing pool equipment is now virtually complete. During this period the EMI magnetic tape decks have been installed and the LSI-11s are now all in use.
Negotiations have been taking place with Dicoll over the maintenance contract for the LSI-11s and this should be completed soon.
Talks have been held with Dicoll and EM I with regard to further purchases this financial year. Quotes have been obtained.
All the grant holders were circulated asking them for suggestions and bids for the new equipment.
A meeting with Prof Coulouris took place to discuss his forthcoming grant application following his recent rejection. As this is likely to be another large proposal it is recommended that a Board of Visitors immediately precedes the Panel meeting to which this application is submitted.
A Board of Visitors went through Prof Grimsdale's application and reported to the CSC through Mr Pearce that a full award should be made.
Dr Gurd of Manchester visited the RL on 5/6 June and discussed the progress of his grant. It was decided to try to organise a SIG in Manchester later this year.
By the date of the DCS Panel meeting the Kirstein Monitoring Panel will have visited DCL again.
No DCS sponsored SIGs or Workshops took place during the period. The coordination team attended the IEE's 'Industrial DCS' Colloquium on 30 May. This allowed us to make several useful connections.
The IEE Colloquium led to visits to the NPL (Demos project of Dr Wilkinson) and the RSRE.
The Academic Coordinator also visited Mr Peter Wilde who is the ESE Committee's SPP in Information & Communications coordinator for Digital Systems Design. This programme is still at the formative stage but Mr Wilde was keen to ensure that the two SPPs cooperated.
Three issues (April, May, June) of the Mailshot were circulated.
The coordination team spent some time preparing the papers for the July meeting.
A set of prospective papers for this conference was refereed by DCS investigators. This activity was organised by the Coordinator.
