The Research Of Hardware/Software Partitioning Based On Particle Swarm Optimization And Simulated Annealing And Clustering Algorithm

With the rapid development of microelectronics technology nowadays, the integration and complexity of SoC(system on chip) increasing greatly. Using traditional system designing method can’t be able to meet the demand of current design. So hardware/software co-design technology has been proposed to solve this problem. While system designers confront a new problem that is how to make the hardware/software partitioning. It demands to optimally allocate the software and hardware modules under the constraints of limited system resources that can meet various design requirements. The study of this problem is of great theoretical and practical significance.This paper first introduces the basic theories and research status at home and abroad of hardware/software co-design. Then it introduces the Particle Swarm Optimization (PSO) and Simulated Annealing (SA). After comparing the respective advantages and disadvantages of two algorithms, it gets the PSOSA algorithm that based on PSO and SA. When PSO updates particle swarms, it uses the Metropolis criterion of SA to increase the diversity of particles. And considering about the increasing scale of the embedded system, this paper introduces the clustering Chameleon algorithm, which has a good performance and effect, to form a hybrid newPSOSA algorithm. It reduces the scale of the problem greatly by clustering, then uses PSOSA to solve the clustering results that can greatly improve the convergence speed and efficiency.In the end, this paper used TGFF to generate random task flow graph and attribute data of the nodes, and made algorithm programs to compare solving performance and quality of PSO, PSOSA and newPSOSA. The experimental results showed that on the whole newPSOSA are better than PSO and PSOSA. Especially when the problem becomes larger, the advantage of newPSOSA on solving performance was very obvious.