Optimization Strategies for Automated Parallelization for Multicore Architectures

Abstract

The article presents an approach to the automation of parallelization based on Petri nets with additional semantic relations. Depending on the architecture of the target computing system, on which the execution of the parallel algorithm is planned, adaptation of the developed parallel algorithm to the existing hardware is required. The paper proposes methods for finding global and local optimum for solving the problem of minimizing the execution time of the algorithm executing on a multicore system. The paper presents the example of application of the proposed method for algorithm adaptation for execution on the processor with two cores.