改进混洗蛙跳算法的软硬件划分方法研究

本文将混洗蛙跳算法应用于软硬件划分，提出一种新型的软硬件划分方法。针对混洗蛙跳算法应用于离散型问题时普遍存在的种群更新过慢、算法寻优方向盲目等问题，本文采用随机步长来改进青蛙种群的迁移行为，采用子种群内进化与全局混洗进化相结合的策略改进盲目全局寻优的情况，并根据无效迭代次数来提前终止迭代以提高算法效率。在划分实验中，改进后的算法的平均最优解比原始算法减小了17.4%～73.3%，平均硬件面积比原始算法大对不同结点数的随机DAG图4.32%～5.81%，平均仿真执行时间只有原算法的42.7%～64.0%。改进后算法在寻优能力和收敛速度上均优于原始算法，可更高效地完成软硬件划分任务。 

Reserch On Hardware/Software Partitioning Method Of ImprovedShuffled Frog Leaping Algorithm

This paper applies shuffled frog leaping algorithm(SFLA) to the partition of hardware and software. It also puts forward a new Hardware/Software(HW/SW) partitioning method.In view of the widespread problems with slow update and ambiguous algorithm optimization while applying SFLA,this paper uses random step to improve the frog populations migration behavior, and also adapts evolution in sub populations and shuffled evolution globally combined strategy to improve blind optimization problem. This algorithm terminates the iteration based on the number of calculated invalid iterations to?improve efficiency. Averagely, the solution magnitude of improved algorithm decreases by 17.4%~73.3%; the hardware area is increased by 4.32%~5.81%; the execution time of simulation is only 42.7%~64.0% of that of the original algorithm.The improved algorithm in searching capability and convergence speed is better than the original one and can be more efficient in terms of completing hardware /software partitioning task.