An Independent Review of the Transputer Initiative that was published as part of the Proceedings of the Closing Symposium at the University of Reading in March 1992 (IOS Press).
This review reports on the awareness and training activities of the SERC/DTI Initiative in the Engineering Applications of Transputers. It presents a factual summary of these activities with some key statistics. It then reviews the subject area in more general terms as a result of interviews with members of the staff of the Initiative's Regional Transputer Support Centres, and others.
In this paper, I report on those activities of the SERC/DTI Initiative in the Engineering Applications of Transputers (the Initiative) which are particularly concerned with stimulating awareness amongst potential users and providing them with training. The activities covered, in order of presentation are:
The Initiative Mailshot; The Transputer Applications; Conferences; Courses given by the Regional Transputer, Support Centres; The Software Exchange Library; Seminars; Workshops; Fairs; Transputer Applications Community Clubs; Standardisation working groups; The Initiative videos and video loan pool; Stands at Exhibitions; The demonstration disk; Support and sponsorship for other bodies and events.
The paper is in two parts, first a factual and statistical summary of the activities including some specific comments on these facts, and then some wider ranging discussion of the perceived aims of the founders of the Initiative in these matters and of how what has actually happened has or has not met these aims.
Many of the things I say in this review have a tendency to be expressed in a somewhat negative manner pointing out minor imperfections in what is being described. Behind all these quibbles is a general feeling that the Initiative has successfully made very many people aware, and has been as successful in their training activities as anyone working in this field at this time.
I acknowledge the help I have received from the staff of the Initiative at the Rutherford Appleton Laboratory and the Regional Centres in collecting the material on which this paper is based. My overall impressions derive both from my long experience of the transputer world from "the inside" when I worked at INMos Limited, from my connections in the World occam and Transputer User Group (WoTUG) and from my recent discussions with people in various organisations both inside and outside the SERC/DTI Initiative. In an attempt to get some input from the real consumers of these activities, readers of the Mailshot were invited to communicate their views to me. The response to this invitation was unfortunately small and there was insufficient time for me to conduct any other kind of "consumer survey" of the activities being reviewed.
A highly visible component of the Initiative's activities has undoubtedly been the Mailshot, which is an A4 size bound document reproduced and distributed at no charge to its recipients. Although originally aimed at UK research and industrial readers, it soon became more widely distributed as is shown by the geographical breakdown of the mailing list in Table 1 below. It would not be difficult to achieve even wider circulation if specific actions were taken to do so.
The first Mailshot was dated October 1987. From January 1989 the Mailshot has appeared monthly (11 times a year) with its distinctive red cover which has made it immediately recognisable on people's bookshelves.
After initial teething problems in setting up a regular pattern of production this has now evolved with the editorial work being done by Initiative staff, and with printing and mailing done under local contracts. It has relied mostly on contributors providing A4 pages suitable for direct reproduction with minimal editorial interference and a commendably short lead time. Early issues tended not to be sufficiently selective in choice of meeting announcements, etc, and suffered from the lack of a page numbering system. Since January 1989 a regular style has been adopted.
Distribution is through a mailing company who use second class post in UK, and accelerated surface post to other addresses. This involves dispatch by air to a central location in each country with onward transmission by surface mail. Worst case delivery within three weeks of dispatch is claimed.
The following categories of contributions have been accepted, and have been suitably grouped within the Mailshot:
Initiative Update (including Regional Transputer Support Centres); Product Update; Courses and Events; Technical Reports; Publications and Job Vacancies.
Initiative Update included regular items listing the key staff at the Regional Centres, news on research contracts, loans, publications, etc.
The most popular sections are the product and course announcements. Companies wishing to announce relevant products or product upgrades are allowed to do so free of charge. However a fee is charged for repeat adverts for an existing product.
It is possible for any organisations wishing to distribute leaflets on any relevant matter to supply these to the Initiative on a bulk basis for inclusion with the mailing. A wide variety of people have made use of this facility, which is charged for at the relevant incremental cost. Leaflets advertising the Initiative's own events are also included when appropriate. The possibility of selling the mailing list to third parties was considered by the coordination team, but the decision was taken not to do this.
Courses advertised have included those given at the support Centres and also a wide variety of other academic and a few commercial locations. Conferences of user groups and other relevant organisations all round the world are also advertised. A useful time-ordered Diary page of conferences and a summary page of Initiative courses are also included on a regular basis.
Individual issues of the Mailshot have been sponsored. In return for a substantial contribution to reproduction and distribution costs a sponsoring company has complete control over a significant number of editorial pages. Companies who have sponsored issues include INMOS, Meiko, Transtech and Dean Microsystems (distributors for Parsytec in the UK at that time).
Technical contributions have appeared on an ad-hoc basis. The Mailshot has not sought a reputation for academic respectability as submitted items are reproduced unreviewed. Technical items thus have a tendency to be limited to papers reproduced from other journals with permission, papers solicited by the Initiative and contributions from people particularly attracted by the short lead time to publication. Authors in more unexpected locations such as Russia have taken advantage of a rare opportunity to get their work seen in other countries.
Job vacancy adverts are charged for. Companies who have taken advantage of this opportunity to reach a well focussed population have expressed particular satisfaction with the outcome, compared to their experience with more general publications.
United Kingdom 2124 Germany 65 Netherlands 55 Belgium 37 France 23 Italy 21 Switzerland 19 Norway 10 Denmark 8 Spain 6 Portugal 3 Greece 3 Turkey 3 Sweden 2 East Europe 78 (mostly Russia, but some in each country) Singapore 73 Australia 56 USA 52 India 51 Singapore 36 Japan 34 South Africa 15 New Zealand 7 Israel 5 Brazil 2 Canada 2 also: PR China, Colombia, Dominican Republic, Egypt, Ghana, Iran, Nigeria, Pakistan, Venezuela Total: 3143 copies to 3034 recipients.
Table 1 shows the geographical distribution of the mailing list at the end of September 1991. This shows 71% to be UK addresses, of whom 52% are "industrial". A slightly greater proportion of the non-UK European addresses are industrial. 158 addresses are of libraries and these are particularly important for displaying posters calling for papers for conferences, etc.
More copies are printed than are explicitly sent to the address list. The additional copies are given away at exhibitions, seminars, etc and are used by Initiative staff as useful free material when performing their awareness activities of all kinds.
January 1989 1300 February 1989 1501 February 1990 2397 February 1991 3034
During 1991 a total of 200 names were removed from the mailing list either by explicit request or as a result of Mailshots being returned by the postal authorities as undeliverable.
An attempt to assess the number of invalid addresses is to be made by sending reply paid cards to a statistically selected sample of names on the list.
There is no doubt that the most visible component of the work of the Initiative to people not technically involved in transputers has been the Transputer Applications (TA) Conferences including the associated Exhibitions.
Three of these have been held as follows:
23-25 August 1989 Liverpool University 11-13 July 1990 Bolderwood Conference Centre, Southampton University 28-30 August 1991 Moat House Hotel, Glasgow
As major international events organised up to a year in advance, it was possible to associate with these Conferences major efforts to get the national press aware of the existence of parallel processing, the involvement of British Industry in this field and the work of the SERC and DTI and their Transputer Initiative.
These Conferences were professionally managed, with selection of papers by a review committee representative of both academia and industry, with specific involvement of the Transputer Applications Community Clubs. Authors of papers selected for presentation are asked to supply a special one page abstract for publication in the Mailshot in advance of the conference and a full paper (of limited maximum length) for inclusion in the proceedings published by IOS Press.
The table of statistics below indicates how successful they were in attracting people as exhibitors and attendees.
TA89 TA90 TA91
Liverpool Southampton Glasgow
Abstracts submitted 55 115 184
Papers presented 53 83 136
(industrial author) 5 12 57
Parallel streams 3 5 4
Exhibitors 36 40 42
Delegates 350 280 380
(industrial) 125 66
The industrial contribution to the technical programme is underestimated in the table as there were several papers with industrial co-authors in addition to those whose authors were all from industrial addresses.
It is also valuable to list the invited speakers, who spoke to a plenary session rather than to one of the parallel streams.
Invited speakers at Liverpool: D May - INMOS N Carmichael - Shell G S Stiles - Utah State University, USA P G Mattos - INMOS M Price - Thorn EMI D Wallace - Edinburgh University D Pritchard - Southampton University F Wray - Topexpress W Leier - Cogent systems, USA J P G Barnes - Alsys Invited speakers at Southampton: C A R Hoare - Oxford University J-M Cadiou - European Commission I R Pearson - INMOS D Shea - IBM, USA P Zimmermann - Volkswagen, Germany H Beduhn - Volkswagen, Germany I Browning - Royal Signals Research Establishment, Malvern C Tricot - Archipel, France A Eppinger - Robert Bosch, Germany L Wiggers - ZEUS Collaboration, NIKHEF-H, Netherlands D G Paddon - University of Bristol Invited speakers at Glasgow D May - INMOS, Bristol G S Stiles - Utah State University, USA A B Watson - Systems Insight Ltd J Mortimer - Intelligent Factories Corporation Ltd, Malvern P D Home - British Rail Research, Derby I Pitas - ESPRIT, University of Thessaloniki, Greece M Yamamoto - Osaka Sangyo University, Japan J Y Lee - POSTECH, Korea V P Bhatkar - CDAC, Pune, India D Stevenson - Parsys.
It will be seen from Table 4 that the organising committee set out to redress the balance of industrial contributors by specifically inviting leading industrial users of transputers to make a presentation. Not all of these talks appeared in the proceedings but nonetheless they were a very important contribution to the events for the people who were there.
A particularly important paper that was not published in proceedings was the keynote address entitled "The transputer and occam: a personal story" given by C A R Hoare at Southampton. This has subsequently been published in the August 1991 issue of the journal "Concurrency - Practice and Experience" and everyone who missed this at the time is recommended to read it now.
At each of the Conferences the opportunity was taken to associate with the main programme a collection of related events on adjacent days. These included both publicly available seminars and tutorials and, private meetings of groups of people who were likely to be gathering at the location to attend the conference anyway.
Industrial sponsorship was obtained to help defray some of the costs. The Southampton and Glasgow Conferences were well supported by local civic authorities. The Glasgow civic banquet, sponsored by the City Council, was particularly noteworthy with musical entertainment included on the programme.
The TA conference series will be continued with future conferences at Barcelona (PACTA) in September 1992 and at Aachen or elsewhere in 1993.
When the Regional Transputer Support Centres started, the preparation and delivery of short courses for industry and other user organisations was seen both as a way of earning income and, probably more importantly, sowing the seeds of transputer expertise in these organisations to pave the way for research and development contracts and collaborative projects suitable for government and/or European Community funding. It was expected that short half or one day awareness courses would predominate, so as to attract a wide range of more senior people. In fact there was a significant demand also for technical courses for engineers, who needed to learn the practicalities of transputer system design and building in order to embark on projects using transputers. The Centres set out to meet this demand.
The tables presented here show the courses that were designed by the Centres. They are based on a review paper prepared by the coordination team in 1990 and on further information provided by the Regional Centres. For each course is indicated the duration, price per student of the course, the number of times the course was scheduled to be given, the number that were actually given and the average number of people who attended. Where a range of prices is shown this reflects a price increase during the period over which the course was offered. In many cases the number who attended was boosted above the number of paying customers by the addition of local academic and research staff from the host university. This was sometimes necessary to avoid running courses with too few students.
Some information on additional courses given for particular clients is included where available. This information is known to be incomplete.
days price(£) sched given ave
Occam and transputer familiarisation course 2 250 8 2 3
Transputer familiarisation course 2 250-330 14 3 3
Transputer appreciation course 1 125 6 0
Technical details of occam and transputers 4 600 6 5 5
C to parallel C 2 280-330 11 6 5
Numerical methods using parallel 2 260 5 4 7
Fortran on transputer arrays in
Fortran C
days price(£) sched given ave
Occam - an introduction 1 120 2 2 12
Occam programming and the TDS 5 360 2 2 5
An intro to transputers applications 1 70 11)
Engineering applications of parallel 1 95 3) 11 14
processing
An intro to transputer software 2 145 2)
Comprehensive 5-day programming course 5 390-495 10)
Advanced programming for transputer arrays 2 195 2)
3L Fortran 1 95 6) 11 5
3L C 1 95 5)
Alsys Ada 1 95 1 0
Exploiting high level languages: transputers 1 125 6 1 11
An intro to transputer hardware 3-4 245-400 7)
Transputer hardware design embedded systems 5 495 3) 5 3
days price(£) sched given ave
Occam - an introduction 5 400 5 2 7
Applying the transputer 1 0 2 2 14
Occam2 for distributed software engineering 3 400 2 0
Management seminar 1 50 3 1
Alien languages (Fortran or C) 3 400 1 0
Basic occam 3 200 2 1 3
Parallel programming methods 3 500 1 0
High level languages on transputers 3 500 2 0
Parallel computing - the way forward 1 0 1 0
Occam programming 3 500 1 0
Parallel program design 2 350 1 0
Southampton were also involved with the CERN summer school at Geneva and taught occam programming to large classes there in two successive years.
days price(£) sched given ave
Transputer appreciation course 1 187.50 8 4 7
Transputer software - technical details 2 357 8 3 9
Transputer software - occam programming 1 187.50 6 2 6
Transputer hardware - technical details 1 187.50 8 3 6
days price(£) sched given ave
Management overview course 1 30 8 4 6
Basic occam course 2 200 7 2 8
Transputer systems using occam 2 200 3 0
Porting alien languages to the transputer 1 100 1 0
Signal processing using the A 100, transputer 3 300 5 0
Parallel Fortran 2 300 3 0
Parallel C 2 300 3 1 7
Introduction to occam 2 199 3 1 8
Parallel processing technology 1 1 1 7
Parallel signal processing 4 1 0
Introduction to parallel C 2 195 3 0
Introduction to parallel processing - a 1 50 2 0
manager's viewpoint
days price(£) sched given ave
Transputer awareness for managers 0.5 0 16 10 9
Introduction to the transputer 1 25 4 1 25
Parallel Fortran 2 250 7 5 8
Exploiting the transputer 1 85 10 4 10
RAL also gave a bespoke course to 2 groups of 25 people at Tessella Support Services plc, Abingdon.
The number of courses actually given, compared with those advertised, was disappointing low.
What does not come over from these consolidated tables is that the successful courses were very highly concentrated into the first full year that they were given, namely 1989. It is particularly interesting to observe that the courses directed at the use of more familiar languages such as For occam to persuade people to attend courses on transputer programming at all? There seems to be a perception in British industry that scientists and engineers do not need to attend external courses, but can always learn new ways of doing things by "sucking and seeing" or "sitting by Nelly".
In 1990 and 1991 the Centres had increasing difficulty in attracting people to their courses, and some Centres took a decision to stop planning them altogether.
On the other hand, advantage has been taken by some of the lecturers involved of their experience in giving courses under the auspices of the Initiative to develop further courses under different auspices (eg COMETT) or to publish a textbook (Northern Ireland).
The Software Exchange library was originally set up at the National Transputer Support Centre at Sheffield. In fact it was the existence of the library which singled out this Centre for additional support and justified the funding by the Initiative of two additional research assistants there.
The library was transferred to the Liverpool Centre in October 1989. At Liverpool emphasis has been placed on the practical goal of getting the contents of the library available to the academic community by on-line file transfer. The library will be taken over by NA Software Limited at Liverpool after the close of the Initiative.
The growth of the Initiative library in terms of number of separately requestable items has been relatively slow. According to published lists the numbers of items in three categories have grown as shown in Table 11.
Compilers Utilities Applications
January 1989 2 9 15
March 1990 2 9 16
July 1991 2 16 27
As a result of their bad experience in getting material submitted to them into a condition suitable for onward distribution the staff of the Sheffield Centre wrote a useful document advising on the effective use of occam and its management with the folding editor. This and a companion article on suitable accompanying technical documentation were distributed to potential contributors and were published in the Mailshot.
The Library Manager at Liverpool felt each item submitted needed an average of a week of his time, to get to grips with. At the end of this time it was sometimes decided that the item was not up to a suitable standard for distribution.
A wide variety of meetings to which potential users were invited were organised by the Initiative and the Regional Centres. These were organised at no charge to the participants. The list of events and activities described below is probably not exhaustive, but shows the types of activities that occurred under this general heading.
A series of free seminars were arranged on behalf of the Initiative by RAL. It had been intended that these seminars would be given at other locations as well as or instead of at RAL, but inertia seemed to inhibit potential organisers from getting their acts together and only the presentation by Sun Microsystems (which was more in the nature of a commercial sales pitch) was repeated at another location (Sheffield).
In approximate descending order of number of participants the subjects of the seminars were:
Windows and Virtual reality; Image processing (given twice); New Software (given twice); Graphics (given three times); Library software; EMR awards; Finite elements; Sun microsystems; Helios operating system; Molecular Dynamics; Databases and spreadsheets
The most popular seminar attracted over 200 people, with an attendance averaging about 70. These seminars have attracted much favourable comment from people who attended.
The Northern Ireland Transputer Support Centre mounted two very successful Transputer Fairs in their catchment area. These were half day events supported by a small number of industrial exhibitors. Financial support was received from British Airways, Ulster Bank, the Industrial Development Board and the Department of Economic Development.
The first was at Magee College, Londonderry, and the second in Belfast. At Londonderry in particular a large number of visitors from Eire were welcomed. The Londonderry Fair attracted over 100 visitors, the Belfast Fairs about 40 each.
This group was established by the Northern Ireland Transputer Support Centre to give all transputer users in Northern Ireland the opportunity to meet for seminars with visiting speakers and to produce a Newsletter.
At Liverpool the North West Regional Transputer Centre set up a local user group called the Liverpool Parallel Club. Organisations are charged £250 to join and there are currently 20 such members. Meetings have been held each academic term since the opening in June 1990, at which invited speakers and members have made presentations. Relevant contributions from supplier companies have also been included.
The Club also produces a Newsletter for members and anyone else interested. The first issue includes an entertaining report on the first visit by Initiative staff to Moscow. A wide variety of product announcements appears in each issue. The fourth issue includes a glimpse at Cray's plans for teraflop processing.
A novel event organised by the Liverpool Centre was a champagne breakfast for leading industrialists. This was held at the local Innovations Centre and attracted local journalists and radio. However the low level of follow-up interest by the industrialists was disappointing.
A series of Workshops were organised by the Coordination Team. The purpose of these was explicitly to stimulate discussion and make recommendations for funding and administrative action to SERC, DTI, etc.
Each workshop was limited to about 50 participants, all of whom spent 2 days at Cosener's House, Abingdon (SERC conference Centre). Participants, of whom about half were explicitly invited by the organisers and half had put themselves forward in reply to advertisements in the Mailshot, were asked to submit a short position paper for circulation in advance of the meeting.
A carefully designed meeting structure was used for all three workshops. This consisted of selected key presentations for most of the first day, closing with a parallel session in which the participants split into 2 or 3 main subject groups to discuss what the principal problems in their subject were.
The second day concentrated on thinking about what to do after the meeting to progress the matters under consideration. This also included plenary and parallel sessions as appropriate.
In December 1987 the subject was "Transputer Development Environments". A report on this workshop was published in the Mailshot dated Feb-Apr 1988. This workshop led to some specific Extramural Research Contracts for items of software seen to be particularly necessary.
In February 1989 the subject was "Applications". The principal outcome of this workshop was the decision to create the Community Clubs discussed in the following section. This workshop can be seen as a seminal component in the process leading up to the decision by the DTI and SERC to follow the Transputer Initiative with a new programme explicitly concerned with "parallel applications". However the eventual form of the programme with a limited number of Centres of excellence was not foreshadowed. The feeling of the meeting was more oriented towards supporting the most promising individual projects wherever they happen to be.
In March 1990 the subject was "Software Standards for MIMD machines" The need for standards in all levels of software was discussed from a variety of points of view, including proponents of the Linda programming model and participants in the efforts currently under way in America to standardise parallel extensions to Fortran which was seen to be concentrating especially on programs for shared memory architectures. Poor portability of programs supplied for the Software Exchange Library was an important motivation for this meeting and the subject of libraries was the topic for one of the subgroups.
There was a strong minority interest in standardisation of occam, and a proposal that this be associated particularly with reliable software and/or embedded systems was made. Shortcomings in occam 2 were raised, but it was felt that standardisation efforts, hopefully with support from the European Community, should be concentrated on the current language and not wait for promised language extensions.
The meeting led to the establishment of two working groups on aspects of standards, discussed below.
Arising out of the Applications Workshop and as an attempt to bring together people from different organisations with common applications interests three Transputer Applications Community Clubs (TACCs) were established.
These were:
No charge is made for membership of a TACC and so their success is dependent on their ability to tap other sources of funding, such as that available from the SERC Computing Facilities Committee and from the Transputer Initiative.
The intention was that these groups would produce newsletters, hold meetings, prepare bibliographies, communicate via electronic mail, etc. Directories including names, addresses, research interests and bibliographies were published by RAL on behalf of all three groups in June 1991. RAL provided a person to act as secretary to each TACC and covered all administrative costs.
The TACCs were invited to nominate representatives to the reviewing committee for submitted papers to TA90 and TA91 and were encouraged to fill one or more complete streams in a section of the parallel conference programme.
Some of the groups represented in TACCs have combined in SERC funded projects to produce demonstrations of transputers in use in their field and to give tutorials in association with relevant conferences.
A hexapodal robot demonstrator developed by CTACC has been featured by BBC TV on the programme "Tomorrow's World" and appears briefly in the introductory credits to the programme.
IPTACC tutorials have been given at several conferences and have combined the tasks of introducing image processing to transputer programmers and introducing transputers and occarn to image processing specialists.
CTACC have also developed a tutorial which was given for the first time at TA91 in Glasgow.
The establishment of MMTACC was an example of the application of Initiative resources outside the field of Engineering applications for which the funding was originally supplied. Funding for MMTACC was terminated when it became necessary to react to reduced overall funding.
Two groups were set up on the recommendation of the 1990 workshop.
The Low Level Working Group (LLWG) considered possible standards in areas such as message passing and resource management. The High Level Working Group had two sub groups, one on the enhancement of Fortran and C for MIMD parallelism and one on the possible standardisation of occam.
All these activities were severely impacted by the SERC funding crisis of 1991. The LLWG have produced a document merging the inputs of four distinct groups on message passing. The Fortran and C work has achieved ESPRIT funding and so will continue in a separate forum. occam standardisation has not yet managed to receive any further financial support.
The first video made by the Initiative used RAL video production facilities with Peter Evans as interviewer and was entitled "Use Your Transputer Initiative". It was completed in early 1989. This 15-minute video has proved very useful for providing some introductory case studies of real transputer projects. Companies featured are INMOS, White Cross Systems, Rolls Royce, British Aerospace and Rockfield Software.
Unfortunately the video suffered from poor sound quality, poor attention to such production details as the appropriateness of the dress of the star performer and poor camera technique of some of the other interviewees. However, it has proved valuable at some of the Transputer Support Centres and for use at events introducing transputers to general technical audiences.
35 copies of the Initiative video were distributed. These have gone to the participating organisations, the Regional Transputer Support Centres, to science correspondents of national newspapers, and to a wide variety of overseas contacts. The Daily Telegraph and the Guardian wrote items prompted by it. Copies are now available in France, Russia, Poland, India, Australia, New Zealand, Japan, Korea, Singapore and Canada. It is estimated that the total number of people who have seen the videos can be counted in thousands.
The new Initiative video now under preparation is being made by a professional production company and will be accompanied by written material and a demonstration computer disk (described below). This video will be targetted more explicitly at industrial and commercial users.
The Initiative has made available 10 copies of their video for loan for a period of up to one month free of charge. Two other videos made by groups in Scotland are also available on loan from the Initiative.
These are a 40-minute video produced by the University of Edinburgh audio-visual services department for the Edinburgh Concurrent Supercomputer Project and an 18-minute video entitled "Parallel Processing and the Transputer" produced by Microelectronic Education Development Centre, Paisley, with help from the National Engineering Laboratory at East Kilbride.
Originally advertised in the August 1989 Mailshot, 17 loans of the Initiative's video were made in the first few months. The tape was then improved by further editing and re-advertised in March 1990. Up to mid October 1991 a total of 154 loans were made. This represents at least 75% of possible usage. Taking into account the fact that not everyone returned the tape on time and that there was usually a short backlog of orders this represents nearly full usage of the available tapes.
Borrowers were asked to state what kind of organisation they were and what kind of group were shown the video. Up to 40% of loans were to industrial companies including British Aerospace, British Airways, British Steel, IBM, Legal & General, Lucas and Siemens. Industrial viewers were usually classified as "company seminar" or "departmental meeting". Academic viewing groups were almost invariably undergraduate, graduate or mixed groups of students. Government research establishments have also been borrowers.
Some of the borrowers have been subsequent users of further services from the Initiative or the Regional Centres.
The Initiative purchased a small professional quality stand at a cost of £4000 for use at Exhibitions of various kinds and it has been used at 21 events. This has been manned by staff from the Coordination Team and/or the Regional Centres who have taken the opportunity to publicise the activities of the Initiative.
This stand has suffered from an absence of real working demonstrations which might have strengthened its impact.
Among occasions where the stand appeared were two PC user shows at the Olympia exhibition centre in London. Unfortunately the idea of the transputer as being relevant to the PC world as an accelerator add-on has never been central to the marketing philosophy. The "Transputer Village" at these shows seemed somewhat out of context and the stand was reported as attracting a "large number of hackers but few serious prospects". However it is probably true to say that all publicity is good publicity and we may never know whether or not some particular transputer markets got their critical stimulus from these appearances.
The Initiative has also appeared as a guest on stands of supplier companies on several occasions, and has attended and exhibited at such disparate events as a Royal Society soiree, a luncheon of the Parliamentary Scientific Committee at the Savoy Hotel, and the UK/LA Arts Fair in Los Angeles.
The coordination team have prepared, with the help of Animated Marketing Ltd, a 16-minute rolling or user-selected demonstration which can be run on any MSDOS PC with EGA or VGA graphics. This is a useful item for use on exhibition stands, etc and includes sections on:
What is a transputer?; Multiprocessing with transputers; Transputer system development; Transputer applications; Transputer architecture; The Transputer Centres.
The Initiative has given support to a wide variety of other organisations working in related fields in Britain and overseas.
Three annual Edinburgh Supercomputer conferences have received financial support, as have two International Federation of Automatic Control (IFAC) seminars, an occam User Group Artificial Intelligence Special Interest Group (OUG AI SIG) Conference and an Institution of Electrical Engineering (IEE) Conference on Software Engineering for Real Time Systems.
The closest analogy to a Regional Transputer Support Centre in another country is probably the Belgian Centre set up with substantial practical help from the Initiative at Ghent. This Centre has not had the benefit of direct government support.
The Initiative has also contributed to meetings of user groups in USA, Japan, Australia and Scandinavia.
Other noteworthy activities have been the involvement of the Initiative with the Polish Transputer Forum and with the Soviet Transputer Association.
When the Initiative started, and the Regional Support Centres were opened, the opportunity was taken to obtain press coverage. Opening ceremonies usually involved the unveiling of a plaque in the presence of representatives of supplier companies, local civic dignitaries, local industry and senior members of relevant departments of the host institution. Table 12 shows the distinguished persons who performed these unveilings.
Sheffield Robert Jackson, Minister of State for Science and Education Glasgow: Ian Lang, Scottish Industry Minister Belfast: Peter Viggers, Northern Ireland Industry Minister Liverpool: Lord Leverhulme Southampton Sir Basil de Ferranti
The Initiative has contributed occasional articles to the trade press and has been instrumental in achieving national press coverage particularly in association with other events such as the TA Conferences.
Conferences and exhibitions have also provided the opportunity for the manufacture and distribution of various items of marketing support trivia showing the Initiative's title, etc. Such items have included pens, document cases, photographs, mounted nonfunctional transputer chips, etc.
In this second part of the current paper I present some relevant opinions on some of the activities enumerated above and also on some general consideration of the Initiative, its Regional Centres and the benefits to INMOS. The absence of specific comment on other activities described above should be interpreted as an implicit favourable review, as any minor quibbles have already been raised alongside the factual reports.
It is appropriate to start with a quotation from a statement of the aims of the Initiative published in its early days. The following four objectives of the Initiative appear in an item by the Coordinator in the October 1987 Mailshot:
Whereas the word "awareness" appears explicitly only with respect to the academic and research community it is implicit in a wider sense within the final objective on technology transfer.
The Mailshot is the cornerstone of the awareness activities of the Initiative. The principal value of the Mailshot has been that it "has been there". Coming out regularly and reliably 11 months out of the 12 each year it could be used as a vehicle for product announcements and advertisements, of courses and other events by all kinds of organisations including the very small. Being constrained to black and white reproduction has made it possible for start-up companies to represent their products in ways which avoid revealing their lack of experience of, or limited budget for, marketing. I have no explicit evidence of the success or otherwise of products solely marketed in this way, but would like to think that some became established as a direct result of their use of the Mailshot.
Being free of charge, issues of the Mailshot could be freely offered to casual industrial contacts by Regional Centre staff. Anyone wanting to remain in touch could then submit a name and address for inclusion in the Initiative's mailing list at no charge and with no need to complete an extended questionnaire.
The fact that it has been reproduced on recycled paper with biodegradable ink may be considered relevant by only a few of the readers, but it does show a concern for such matters by those who produced it which is highly commendable.
Attempts are currently being made to ensure the future of the Mailshot after the other activities of the Initiative have closed down. Perhaps a more exciting name could be chosen for the successor publication?
Most people who had attended the TA conferences have considered them a great success. Being particularly successful in attracting exhibitors, and being held in spacious and appropriately organised accommodation, made the conferences ideal for the informal interchange of views which are usually deemed to be of the essence of such meetings.
Mixed feelings have been expressed on the details of the technical programme. The invited papers were mostly of very high quality, but with the occasional tendency for some English speaking presenters to forget that they were at an international event and ought to slow down a bit! The contributed papers included some which a different reviewing strategy might, with benefit, have eliminated. The Glasgow limit of 6 printed pages tended to make some of the printed papers too short to contain adequate presentation of their subject matter. By reducing the number presented and so the need for an excessive number of parallel streams, it would have been possible to allow a more generous allowance of space in the proceedings for those who needed it. At Southampton one of the five streams managed to attract only 6 people to one of the papers. The time for oral presentation was also considered inadequate for many of the subjects presented. However, the prime purpose of a conference paper is to alert the reader to the existence of the work being reported. Anyone further interested can then make direct contact with the author. The TA conferences have definitely been successful in this respect.
The success of the exhibitions was self-perpetuating as the series got the reputation as being the most effective place for companies to put on their best shows. It was particularly encouraging to see rival processors being actively promoted in a transputer context for the first time at these events, with Intel i860s appearing at Southampton and the Texas Instruments S320C40 at Glasgow. Texas Instruments paid the Initiative a significant compliment by choosing TA91 for the European launch of this product.
At Glasgow it was decided to use a major hotel as the conference base, rather than an academic location. This gave the opportunity to offer a high standard of accommodation to those who wanted it, but at a correspondingly higher cost. Whether this was a good idea or not depends very much on who one talks to. Even those who opted for university accommodation and free bussing between locations had to pay more than twice what they had paid at Liverpool two years earlier. Although many people chose not to attend this meeting, the fact that numbers were increased over the previous year, in a more remote location, shows that the right decision had been taken.
Courses have been of two distinct types. Short courses of one day or less, sometimes free of charge, addressed at managers in an attempt to make them aware of the potential of parallel computing, and longer technical courses directed at the engineers who are actually going to design and implement transputer hardware and software systems. These types match the two components of the subject of this paper.
The value of an awareness course is very dependent on the make-up of the group of people attending. The reason for this is that the impressions made by other members of the group are just as important as those made by the lecturer. Perhaps mixing academic and industrial people on the same course may not always be a good thing, as their viewpoints are likely to be too distinct to benefit each other? On the more technical courses the opposite probably applies and the interplay of views from academic and industrial students on these contributes positively to the process of technology transfer.
It has not been possible for me to survey course attendees in any depth, but the impressions that have been obtained have mostly been entirely favourable. The style of presentation in technical courses has been perceived in a few cases to be too close to undergraduate teaching, but in most cases has been seen as entirely appropriate.
Particularly valuable has been the written course material which students took away with them, some of which is of a high standard, and the hands-on experience of transputer software development, especially with the easy-to-use Transputer Development System (TDS), included in all the longer courses.
Originally the coordination team had hopes that courses developed at one Centre could be packaged and delivered at other Centres. Unfortunately there was no consensus on how, or even whether, this should be achieved. One problem was that all the Centres wanted to get going at approximately the same time, and as the subjects were exceedingly novel each lecturer had to go through his own learning process very close in time to delivering his first course. Having done this it is not surprising that each lecturer felt that he would be most comfortable presenting the subject in a way that he felt to be right, rather than wait for someone else's ideas with which he would be less in sympathy. Some lecturers said that they would have found it valuable to have seen other Centres' material at the time they were devising their own courses, but that in practice that could have imposed unacceptable delays of many months.
Five Centres eventually collaborated to differing extents by exchanging course material and then negotiating the rights to use it on an individual basis. The remaining Centre took the line that it was not prepared to surrender the intellectual property invested in its courses without any financial recompense. In the circumstances this was a natural reaction to pressure to become more business like in their activities. Other Centres saw the business benefit of sharing their experience with each other as outweighing the alternative viewpoint.
Various reasons have been put forward as to why the demand for courses seemed to dry up in 1990. The general economic situation is probably one valid reason, manifesting itself both in a general reluctance to spend money on training and especially to consider novel technical solutions which are perceived as inherently risky. Another set of reasons concerns the perceptions of engineers that there is no longer anything especially difficult to learn as the ideas of parallel processing have become more commonplace and transputers can now be programmed in Fortran and C "which we already know" anyway. This latter viewpoint is disappointing as there is a lot to learn about the design of parallel software and the structures of parallel programs in general which would have been covered on such courses.
When the demand for course places dropped below the level of viable group teaching, several Centres continued training activities on a one-to-one basis for people prepared to pay for this. The level at which it was possible to charge for such tuition probably restricted it to loss-leader situations to attract new industrial partners. In terms of spreading the word to a greater number of people the more valuable activities were probably the giving of specially tailored courses on client sites. Some of these are noted in Tables 5-10, but there are probably several more that have not been brought to my attention. Some companies in particularly sensitive applications fields are particularly reluctant to send staff to open courses. For these companies in-house training is probably a better option. In-house courses given by staff of the North West Regional Transputer Support Centre have been particularly successful.
In 1989 the National Transputer Support Centre and the Training Agency produced a report on the needs for training in parallel programming. One of its most significant recommendations was that without a major initiative on the training front Britain would be in serious danger of being overtaken by the rest of the world in the next 5 years.
The aims of the Software Exchange Scheme were identified in the report of the SERC CFC Working Party of October 1986 as follows:
This programme would be expected to produce software packages of general community interest and relevance.
To capitalise on this a Software Exchange Scheme is proposed whereby such software products would be collected together in a special Software Library maintained by the National Transputer Support Centre. Each product would have to be documented by the originating Group as a condition of the loan of hardware and software from the Loan Pool.
Such packages should have a declared support level (provided by either the originating Group or National Transputer Support Centre) and should be licensed for use by Academia at no cost.
This aim has been extended to include software specifically identified as deliverables of the Initiative's extramural research contracts. (See M Sabin: The EMR Programme). In addition, the opportunity has been taken by a variety of transputer programmers to contribute to the library some generally applicable tools that aid the development of transputer software.
In practice the most frequently requested single item in the library has been the Origami folding editor. This was produced, on the side of his official work, by a user in an industrial environment not specifically in the software tools business. An important feature of the Origami editor is the fact that it is an ordinary DOS program for the PC. It also has the ability to read and write the text file format used in the TDS.
An unfortunate fact about many programs in the Software Exchange Library is that they are closely dependent on a particular target configuration and/or development environment. It is unfortunate that the INMOS TDS, as well as introducing a text file format that is non-standard, has also been unable to lay down sufficiently rigorous conventions of its own to ensure portability of programs between its own different versions, including the two principal INMOS versions and those offered by Meiko for their Computing Surface and by Cresco-data for the Apollo workstation.
The OCpro TDS editor enhancer is an example of a tool tied to a particular development environment. This tool, originating from Salford University, was considered by Sheffield to be the most valuable item in the library at the time.
On the applications side the most popular item has been the FFT algorithms, reflecting the fact that signal processing is one of the most widespread applications of transputers.
Except where there is a dependency on special purpose hardware, reasonable portability at the level of occam source code is usually possible, but many items in the software exchange library are not distributed as occam source and so this potential is wasted. It is unfortunate that source was so often withheld from distribution by the originators as a major benefit that could have been obtained, namely a body of code that could be read to generate ideas on coding and structuring style, was not possible.
There is a fundamental problem in the setting up of software libraries that the software community have not yet come to terms with. The fact that software costs roughly the same to write whatever it is to run on, but can be replicated very cheaply, means that users of common platforms such as the PC or Unix have an advantage over everyone else in terms of availability of software at very low prices. We need to find a way to reward the writers of software for more unusual machines so that they are not inhibited from giving their software wide distribution at a price familiar to the end user.
Maybe the concept of "shareware", which is now taking hold in the PC and Unix communities, could be relevant here?
The Initiative was set up as a result of pressure from SERC's Peer Review System, which perceived a large number of bids for research money involving procurement of transputer systems. What these researchers wanted was hardware and software to enable them to enter the new world of parallel computing at an early stage.
Because of the opportunity presented to promote an industry segment in which British companies, especially INMOS, were at that time dominant, the Department of Trade and Industry also became involved. DTI, in particular, perceived some good products which were not being as successful as they could be, and so joined the SERC in the Initiative in particular to promote the awareness in potential industrial users of the subject of parallel computing in general, and transputer systems in particular.
A central coordination team was seen as an essential component of such a programme. It was obviously correct that the coordination team was not located at a particular academic site. Pressures within universities and polytechnics are not always compatible with policies to collaborate on a nationwide basis. Coordination was therefore based at the Rutherford Appleton Laboratory.
It is unfortunate that resources were not available at the start of the Initiative to help the Centres get established with a visible commonality of style. If this had been achieved it would have been to the benefit of all concerned and might have avoided some of the conflict towards the end of the Initiative when Centres have been put into a position where they have had to compete with each other both for industrial contracts and for a share of the next round of government money.
A valuable series of training courses has been laid on at the Initiative's expense recently for staff of the Regional Centres. These have been on Marketing and Presentation Skills, Marketing Documentation, Project Planning and Business and Financial Planning. Some people have criticised these as being too expensive, but on a per capita basis they appear not to be much more expensive than the technical courses given at Regional Centres. They have been given by professional business training companies and have included both pre-course and post-course documentation tailored to the particular audience, as well as three days in the classroom.
The value of these courses would have been enormously greater if they had been given at the start of the Initiative. Particularly obvious when observing the leaflets, etc, produced by the Centres is the lack of any house style. Also, the value of getting the more junior members of the teams at each site to meet their opposite numbers should not be underestimated. It seems that the only other occasions people from different Centres got together on Initiative business was at the formal meetings of the Management Group, at Centre Directors' meetings and at Centre Managers' meetings.
The Regional Centres were established with the hope that they would become self-funding after three years. Initiative funding was later extended to a fourth year for some Centres. A wide variety of ways forward has been found by the various Centres.
With the Initiative only funding one and a half posts at each Centre, and at least a similar number committed by the host institution(s), the style and success or otherwise of the Regional Centres has been disproportionately dependent on the contribution of the host institution. Each Centre has had an academic director, usually a professor. It is in the nature of universities that professors have strong departmental loyalties, and the styles of the Centres have been very dependent on the particular subject department in which the director is based, and the nature of his or her relationships with other heads of department (usually Computer Science or Electronic Engineering) with an obvious interest in the Transputer Centre. At some Centres the "other" subject has a close relationship, at others a very tenuous one. At the Centres with two host institutions the potential difficulties are multiplied. In one such case these were successfully surmounted.
All these considerations lead to what one observes in visiting the Centres, that the personalities of the actual director, manager, other technical staff and administrator are exceedingly important. It is not an accident that one of the most successful Centres is the only one where the same manager has been in post since shortly after the Centre's foundation. It is also significant that this manager (and no other I can identify) had previous experience in a managerial role. One of the Centres had a succession of academic managers on yearly secondments. Each of these managers made an excellent contribution, but where was the continuity? It is unfortunate that, after the person originally appointed manager of the London and South East Centre moved on, a full time successor was not appointed so that a separate identity could be maintained from the coordination team also situated at RAL.
The physical surroundings in which a Regional Centre is located can have an effect on the initial impression on a potential industrial client. An attractive modern Science Park can make it easier to look clean and tidy, but is it always worth the rent compared with a spacious but antiquated set of rooms in a Victorian University?
A feeling expressed by people at several of the Centres is that they were given an impossible task to perform. Particular difficulties they have mentioned have been the arbitrarily moving goalposts over the years, absence of sales and marketing skills amongst staff, inadequate travel budget, the top heavy local management committees, the difficulty of a small team to present an impression of expertise in all subject areas. There was also much adverse comment on the immaturity and lack of formal maintenance arrangements for hardware and software products being used.
The success or otherwise of the Centres in attracting significant industrial contracts has been variable. Very important has been the pattern of existing contacts already set up by the academic director. Given the way academic credit is nowadays believed to be a quantifiable commodity, it is not surprising that in some places it is seen as better for industrial contracts to be set up with existing academic departments rather than with a Transputer Centre which can appear peripheral to the host institution as a whole.
There was a mistaken perception in some Centres that there was pressure for potential industrial contacts to be set up through the Coordination Team rather than by direct local actions. This was compounded by a perception that companies should be directed to a particular Centre on a geographical basis, where a subject-based decision would often have been more appropriate. There was something to be said for allocating Centres on a geographical basis in the first instance, as it is especially appropriate for courses, but as the Centres got established the polarisation on a subject basis which was inevitable because of the small team sizes, could have been given more acceptance and even encouragement from the Coordination Team.
This was an area where the existence of central coordination was perceived as a negative feature by the Centres, especially when comparing themselves with Centres set up in different ways such as those at Edinburgh University and at Bristol Polytechnic, both of whom enjoyed much greater self determination.
The Initiative was not set up specifically to benefit INMOS, but as this is a Transputer Initiative, and the transputer is an INMOS product, it was instructive to enquire of INMOS how it perceives the influence of the Initiative on its present and future activities.
INMOS acknowledges the importance to it of the awareness activities of the Initiative, which have undoubtedly extended their markets, and it has participated in the activities in many ways. INMOS also helped in the establishment of training courses by involving their own training staff in early courses at RAL and by the provision of material.
In the present financial climate INMOS is constrained to put most of its marketing effort into sectors which it sees as likely to give the maximum return in the short term. These are seen to be a fairly small number of hopefully very high volume embedded systems, relatively few of which will be designed and manufactured in Britain.
The question then arises as to whether the Initiative is addressing researchers and developers whose ideas will evolve into such high volume products in a short time scale. INMOS currently believes not, but on the other hand it also appreciates the value of establishing a constituency of potential customers who are at home with the transputer and the view of parallel programming that it encourages and supports. This is where INMOS perceives the value of exercises such as the Initiative and is particularly keen to encourage the continued existence of Centres of transputer knowledge and also eventually their expansion into other countries. INMOS' parent company, SGS-TIIOMSON, is already expanding its activity in Eastern Europe and the influence of the Initiative in Russia is seen as potentially very valuable in the longer term.
INMOS is concerned to keep the transputer constituency active and hopefully self supporting. Unfortunately, however, a wide customer base does not necessarily equate to high sales volume, and it can be expensive to support a large number of relatively small customers. This is the reason for the INMOS policy to direct all sales to such customers through the SGS-THOMSON network of distributors. It is therefore particularly important for the staff of such distributors to be adequately trained for their role in supplying and in enhancing the market.
The Initiative can take much of the credit for an increase in awareness of the transputer and parallel processing, and is particularly to be commended on the success of the monthly Mailshot and the annual Transputer Applications conferences, both of which are likely to survive the end of the Initiative.
Reasons for poor take up of the training opportunities can be found in a variety of circumstances, including the immaturity of the technical field and especially of software development products, the low priority given to training by British firms in times of financial stringency, etc.
A missing component of the broad package that could have been associated with the work of the Initiative, by additional funding from the appropriate departments, is work towards the inclusion of parallel processing in undergraduate and even in school levels of education. Parallel processing is here for ever; the sooner the engineers and programmers of the future start to appreciate the concepts the better. The students of today are the decision makers of the future. The Initiative should not be blamed for this particular shortcoming.
It was a brave decision to associate an Initiative of this magnitude so closely with the products of a single company (by including the word transputer in its title). In retrospect I think that it was an error to use the word so explicitly; the same work could equally well have been done under a more general title without changing the de-facto situation that most hardware bought by the Initiative would still have been transputer-based.
The training and awareness activities of the Initiative have shown the need for standards in the new area of parallel systems. It has been necessary to base courses on whatever software systems were available, rather than any widely accepted system with a well established following. The Initiative has been successful in getting appropriate extensions for scientific languages onto the international agenda. It is hoped that when the need for a language, accepted by international standards bodies, which really addresses the problems of verifiable and reliable software systems is finally accepted (let us hope not as a result of a major catastrophe resulting from inadequate software!) then occam is still around to step into the gap.
It is important that the work of the Initiative on standards is guaranteed continued public funding. This work has started to address the problems, arising from the immaturity of the field, of designing really portable parallel software. Remember that it is not long since Unix was considered immature!
In a sentence, the conclusion of my review is that awareness of parallel processing and of the transputer in particular has undoubtedly increased substantially during the life of the SERC/DTI Initiative, but that the community did not take sufficient advantage of the opportunities for training that were provided.