Barcelona, the newest Olympic city, was the venue of PACTA '92 - Parallel Computing & Transputer Applications 1992 - Conference. This event brought together many of the leading researchers and companies representing the state of the art in parallel processing.
183 papers were presented, and showed the emergence of parallel processing as a fundamental tool in many fields of computing. Parallel sessions included: Algorithms, Robotics/Control, Scientific Computing, Linear Algebra, Signal Processing, System Tuning, Neural Nets, Simulations, Image Processing, Finite Elements, CFD, Transputer Applications, OS and Databases.
In addition to the technical presentations, six keynote lectures were presented by invited speakers. These talks provided a wider perspective of the current developments taking place to promote and engender the exploitation of parallel processing. A review of the High Performance Computing Initiatives being setup in both Europe and the USA was presented. The remaining talks featured Finite Elements, Weather Simulation, Oil Production and Automatic Parallelisation.
The general focus of the technical presentations was the "Teraflop Grand Challenge" - construction of a parallel supercomputer capable of achieving teraflop rates of calculation.
INTEL, IBM, CRAY, Thinking Machines, Fujitsu and Convex each presented their view of what a teraflop machine would be. It was interesting to note that based on current technologies a teraflop computer would cost hundreds of millions of pounds; need to be housed in a small warehouse; and require an extensive cooling system. To operate the machine would require upwards of three Megawatts, requiring its own electricity sub-station to down-load off the national grid. It was noted that switching such a machine on or off would require the approval of the local Electricity authority on each occasion! There was much debate on who could afford to run a teraflop machine and what this would mean to the engineer.
Running in parallel with the conference sessions was a large exhibition of the latest products in the parallel field. The products displayed ranged from large-scale supercomputers (Cray and Fujitsu) to workstation clusters (IBM) down to the current generations of chips from Texas Instruments (C40), Intel (i860) and INMOS (T8OO and T9OOO). It was fairly obvious that the existing transputer range is no longer reigning supreme amongst the stands and that the C40 is currently leading the way, with the i860 close behind.
A panel session on the final day considered how to move applications to teraflop architectures. The panel expressed the need for conformity across the developing architectures to enable easy migration between systems. The panel also stressed the need for the training of engineers to learn the mechanisms required to build and generate code to operate on large-scale multi-processing systems.
Overall, the conference was held in an exotic city and its seemed an appropriate location to discuss the increasingly exotic nature of parallel processing and its future directions.
PACTA '92 was the first time that memetic algorithms were discussed in a European Conference:
A *memetic* approach for the traveling salesman problem: Implementation of a computational ecology for combinatorial optimization on message-passing systems, P. Moscato and M.G. Norman, Parallel Computing and Transputer Applications, Proceedings of PACTA *92, Barcelona, Spain, Sept. 1992, vol. 1 (pp.177-86), IOS Press, Amsterdam (1992), ISBN 9 051 99096 0.
This is one of the most-cited papers of the conference.
