论文部分内容阅读
软硬件划分是软硬件协同设计中关键步骤之一,并且随着设计复杂度的增加,逐步成为一个具有挑战性的优化问题.提出一种基于从众和声粒子群算法(conformity particle swarm optimization with harmony search,CPSO-HS)的并行软硬件划分方法.按生物行为学理论,个体粒子具有从众行为,趋向于靠近群体粒子聚集的安全地点,以避免被捕食者袭击.CPSO-HS算法通过模拟这种从众行为,能够保持搜索种群的多样性,以避免陷入局部最优,有利于逼近全局最优点.通过改进和声搜索算法(harmony search,HS)的初始化策略,将HS集成到CPSO-HS中,在当前全局最优解附近提高算法的搜索精度,有利于提升解的质量.以上两步的有机结合,增强了CPSO-HS算法搜索的多样性和集中性.进一步考虑软硬划分方法的特殊性,其中最耗时的过程是计算软硬件的通讯代价,因此在常用的PC平台上采用并行策略加速该过程,以便在大规模的软硬件划分问题中有效减少整体运行时间.最后,通过基准任务测试集验证了本文方法的有效性.
Hardware and software partitioning is one of the key steps in collaborative design of software and hardware, and it gradually becomes a challenging optimization problem with the increase of design complexity.A new algorithm based on conformity particle swarm optimization with harmony search, CPSO-HS) .According to the theory of biological behavior, individual particles have a herd behavior that tends to gather near a group of particles in a safe place to avoid attack by predators.CPSO-HS algorithm simulates this The herd behavior can keep the diversity of the search population and avoid falling into the local optimum, which is good for approximating the global optimum. By improving the initialization strategy of harmony search (HS), HS is integrated into CPSO-HS, In the current global optimal solution to improve the search accuracy of the algorithm will help to improve the quality of the solution.The organic combination of the above two steps enhances the diversity and concentration of CPSO-HS algorithm search.Further consider the particularity of the soft and hard method of division , One of the most time-consuming process is the calculation of communication costs of hardware and software, so the common PC platform using parallel strategy to speed up the process, in order to The large-scale hardware and software partitioning problem can effectively reduce the overall running time.Finally, the validity of the proposed method is verified through the benchmark task test set.