Transputer and its application in the mining and process industries – a multiprocessing perspective

"General purpose uniprocessor systems used in control applications, however fast they may be, at a certain stage attain saturation, forcing the user or designer to consider other avenues for solving the problem. This paper gives a broad overview of multiprocessing with special reference to transputers. Virtually all aspects, from parallel systems real-time control to artificial intelligence (AI) and expert systems, have been highlighted. The application of transpurters in unilayer and multilayer control are discussed in detail, with specific attention to sensor integration. Two case studies describe at length the utilization of transputers. IntroductionIn conformance with the high level taxonomy of parallel computer architectures as given by Ralph Duncan(l) (an improvement to Flynn's classification, which maintains the essential elements of useful taxonomy based on instructions and data streams), the MlMD architectures are categorized into two groups: (I) shared memory, and (2) distributed memory type. While the shared memory multiprocessors are more popular than the message-passing multiprocessors because they can be programmed in a conventional sequential fashion, their aspect of the architecture creates a perfonnance bottleneck. In contrast, message-passing multiprocessors (MPM) suffer from no such limitation and can attain linear speed increases that are directly proportional to the number of processors put into use. The MPM implementation also has its own advantages and disadvantages. Because the performance of the system is heavily dependent on it, the interconnection strategy in a distributed memory MIMD machine is of paramount importance. The transputer(2.J), a multiprocessor node of the distributed memory type, is becoming very popular. The advantages offered by the transputer make its application in the mining and processing industries very promising."