Eurographics 80

Eurographics 80

:~ ~ ~!~~zi~~ ~:i~?i ~~i~i~ ~ recognition of the continuing pattern of change in the nature and composition of the labour force. This theme serv...

271KB Sizes 1 Downloads 58 Views

.....:~ ~ ~!~~zi~~ ~:i~?i ~~i~i~ ~

recognition of the continuing pattern of change in the nature and composition of the labour force. This theme served as a convenient link and starting point for the second round table, on education and industrial training, by demonstrating the urgent need for education and training to provide the awareness and understanding and also the new skills now seen to be in short supply. Panel members and participants seemed to feel that formal education in all countries had been overtaken by technological developments and is now failing to match the changing requirements of industry and society. In discussion, the difficulties of introducing changes to existing courses and syllabi, also lack of suitable courses, were stressed. Points were also made on the need to educate the educators and ensure that CAD was introduced as a tool in the various disciplines. As far as industry was concerned, the functional requirements and job descriptions in the new technology seemed able to be defined, but matching them to educational courses and devising suitable course content was more difficult and contentious.

Industrial training Problems of industrial training provoked lively comment. Managers as a class were considered to need special attention, but practical difficulties in providing courses of awareness and understanding were thought by some to be too great to be overcome. Others thought differently and cited evidence of 2 -3 day courses which had been considered successful. The dividing line between formal education and state-sponsored training, whose responsibility it was for industrial training and how the cost should be borne was subject to debate but no clear conclusion. Industrial representatives accepted responsibility for on-the-job training and cited the cumulative way in which technology transfer could be achieved by initial training of one man in a firm who then in turn trained others and so on. This suggestion met with objections

42

from others who pointed out that whilst this technique was feasible for large firms it was not possible for the small firms. It was concluded that the answer probably lay in positive action by groups or association of firms in industry to cooperate in training and interchange of information - preferably with some government help to set up information centres. This provoked further suggestions for support and cooperation on a European scale, such as might be effected by the Social Fund of the EEC. To effect an improvement on this or any other related aspect, the suggested remedy was more effective briefing and lobbying of national representatives on key priorities.

~ i~

Overall impressions were that MICAD80 could be judged a success and had justified its aim of briding the gap to the smaller firms. The advantage of holding MICAD-80 within an enormous exhibition centre and having a parallel set of CAD demonstrations on exhibitors' stands within SICOB was offset by the difficulty both of finding them and also of finding the time since every day had a very full programme of parallel sessionswith no gaps or free time. No doubt further thought will be given to these points when consideration is given to staging a follow-on event in two to three years' time.

Arthur Llewelyn

Eurographics 80 3-5 September 1980, University of Geneva, Switzerland, Organized by Eurographics Association Eurographics has been established as a Foundation in Switzerland and registered as exclusively for educational and scientific purposes, the announced aim being to cover theory, design, implementation and application of computer-generated graphics and interactive techniques and facilitate man-machine communication and understanding. Eurographics 80 was its first official conference. The enormous expansion in creative computer usage which graphics could bring was perceived early in the 1960s when its potential for manmachine communication was first demonstrated. Its realization, however, had to await slowly acquired experience in applications and software and systems design. Disappointment at the slow progress caused graphics to be regarded for many years as a gimmick. Now developments in software and hardware and, above all, feedback from application experience resulting in improved system capabilities have restored graphics to its rightful position of prominence. The renewed surge of interest in graphics as an essential building block in systems design and a user

tool was demonstrated by the contributions at all levels to Eurographics 80. This was evident in the program content which, although showing signs of the usual difficulty in structuring, such a diverse set of papers, certainly succeeded in uncovering some new work on man-machine interaction and graphics software as well as interdisciplinary applications. This had the effect of widening the circle of participation as was obvious from the questions asked and issues raised in discussion all too familiar to the 'old hands' but symptomatic of the extremely slow diffusion of knowledge outwards from centres of expertise. The perennial problem, evident in this conference as in others, lies in the enormous range of applications which can be covered by any multidiscipline tool such as graphics and the dilemma posed by brining together various levels of interest from expert to would-be user. This wide audience spectrum was mirrored in both the papers presented and the discussion. As the conference unfolded, it was apparent that many people want to do creative things, are groping for understanding, and feel frustrated and impatient both with the lack of good information (required in simple terms and not technical jargon) and the clumsy interfaces they must use[ One

computer-aided design

. . . .

~ . . . . .

. . . . . . : :

::.::::.:: =

. , .

:.:.:........

:==

.:,

:::.:::::

.......

: ....

becomes conscious of the great gap that still exists between such users and would-be users and the hardware and sbftware designers and manufacturers. If Eurographics can help to bridge this gap by articulating their needs and bringing about an improvement in user software tools it would indeed more than justify its existence. The difficulty in effecting a real improvement may prove to be the problem of bridging the gap between the experts (however defined) and the commercial world of manufacturers and the growing number of purveyors of software systems. Economic pressures exerted through the rising costs of

:.:::.:.:

:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

============================================

:;.:

:

;

.

.

.

.

.

.

.

.

.

.

.

,

.

.

.

.

.

.

,

.

. . . . . . . . . . . . . .

'going it alone' and need to reduce costs of training, upkeep and enhancement and access information banks may all force a trend toward common 'system interfaces and data formats but this is a very slow process and progress in this direction may well await a more positive commitment from the software service industry. This theme and the design methodology for a common systems approach formed the core of an invited paper by William Newman who drew attention to future trends and the problems which would be created if some agreement on common data formats, user conventions, and system interfaces

. . . . . . . . . .

. . . . .

. . . . . . . . . . . .

..

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

were not adopted. The logic of the argument was inescapable and in a sense presented a challenge to the Eurographics Association to exert their influence and do something positive about the existing situation which threatens to get worse as the area of graphics application expands. The question is can such a body act as a focal point for the many disciplines involved and provide unifying guidance for future developments? There must be many who would wish the Association well in such a task.

Arthur Llewelyn

17th Design Automation Conference 23-25 June 1980, Minneapolis, MN, USA. Sponsored by A CM/SIGDA and the IEEE Computer Society DA TC. Despite the well attended architecture and mechanical engineering sessions, the Design Automation Conference is still mainly about electronics. This year's talking point was VLSl (very large-scale integration) - defined by Car[ McCaw of IBM Hopewell Junction as chips containing 50 000 to 500 000 logic circuits beyond RAM storage and, according to Lawrence Rosenberg of RCA Laboratories, one to two orders of magnitude more complex than our design automation systems can comfortably handle today. Indeed it is this complexity of design that is the biggest barrier to a comfortable continuation of the familiar exponential growth curve, shown at every single 'miracle microchip' presentation for years, that doubles device count every two years or so until the physical limits of 0.25/am or so line widths are achieved. That 'complexity' has real mathematical meaning was demonstrated in two invited papers, by S Sahni of the University of Minnesota and W E Donath of IBM Yorktown Heights. They explained clearly the buzz-term that even now is drifting into the DA professional's vocabulary: NP-complete. NP means nondeterministic polynomial and is used to describe a class of problem, like that of the famous

volume 13 number I january 1981

travelling salesman trying to plan an energy-efficient journey through various cities. For each additional city, the time needed to try every solution - the only way to find the shortest route - increases explosively. The problem can only be solved in polynomial time using the theoretical, and so far mythical, Turing Machine. The complete in NP-complete means if one problem can be solved, every problem in NP is soluble in polynomial time. Most DA problems are like this: placement, interconnection and routing, test-pattern generation, fault detection. Under the worse-case, these problems are intractable, and the same is mostly true for 'close' solutions. A glimmer of hope could be in highly parallel algorithms running on specialpurpose array processors. These problems were further emphasized by David Slate's entertaining but apposite conference-lu nch talk 'Computer chess Northwestern style'. In the 'Design automation and VLSl in the 80s' panel, Benjamin Lee of Calma proposed the more realistic, and old-fashioned, solution - design assistance rather than total automatior~ The designer uses symbolic graphical circuit descriptions, stick diagrams, plus auto-compaction consistent with design rules, but uses his own brain for the rest. This view was confirmed by Carl McCaw's questionaire put to the nine panellists, which also addressed the silicon utilization versus design-time

problem. Most panellists agreed that more use will be made of standard functional building blocks - faster turnround and lower design cost at the expense of some silicon area. Testability too, said Sam Daram of Hughes Aircraft, will also take up space. Testing of course was one of the other major challenges identified by Larry Rosenberg in his invited paper. It received an invited paper/ tutorial of its own from A W Nagle and J Grayson of Bell Labs, Holmdel. This text should prove invaluable for some time to all those to whom it is addressed: the designers who are not yet convinced of the necessity of design for testability. Some basic circuit features make digital test development very difficult and circuit packs with LSl components have the problem that simulation models may not be available, either because of complexity or because schematic information is not available from the manufacturer. But if it is difficult to test a board containing ICs, what of the ICs themselves? The authors demonstrated how difficult it is to detect a single stuckat fault. In the main sessions Miron Abramovici of Bell Labs Naperville described an effect-cause analysis which does not need fault enumeration or a comparison between expected and actual response. The circuits can be modelled at gate or functional level, can be redundant and need no initi-

43