Abstract
This paper proposes a method of reducing the overhead and improving the speed of the Petri netβcontrolled multiprocessor system. The authors have previously proposed a Petri netβcontrolled multiprocessor system in which direct control of such operations as synchronization and exclusive control can be realized, once the parallel processing is described by a Petri net. When a microprocessor is employed as the controller, however, a large control overhead is required and the processing speed is not improved even if the highβspeed processing is considered by the parallelism. In the method described here, speed is improved by a search for firable transitions, the elimination of places that have been made superfluous by return or kill operations, the use of a functional memory with a large word width for processing global places, and the implementation of the controller in hardware. The overhead of the proposed system is evaluated by a simulation, which indicates that operation will be two orders of magnitude faster than in the previous system. The overhead ratio is small even for smallβscale processing in which the process unit (granularity) is the statement and that parallelism is thoroughly utilized.