Execution of real-time expert systems on a multicomputer
E. Bahr, Franz Barachini, J. Doppelbauer, H. Gräbner, F. Kasparec, Thomas Mandl, H. Mistelberger · 2002
A transputer-based architecture, the parallel PAMELA research engine (PRE), has been developed to support the needs of the parallel version of PAMELA. The mapping of the fine-grain algorithm onto the coarse-grain PRE architecture is discussed, and preliminary performance data, including benchmark run-time measurements, are given. The suitability of the proposed matching algorithm for execution on a machine like PRE is demonstrated. It is shown that the scheduling algorithm is able to manage the load-balancing problem, resulting in reasonable utilization of processing power and a high speedup for most benchmark programs investigated. The parallel PAMELA system can be easily integrated into existing CIM architectures.>