基于改进遗传与模拟退火融合的RISP软硬件划分

Hardware/Software Partitioning of RISP Based on Combination of Improved Genetic Algorithm and Simulated Annealing

摘要:

软硬件划分是可重构指令集处理器在软硬件协同设计中的关键问题,通过对比遗传算法和经典模拟退火算法的优缺点,提出改进遗传算法的适应度函数,同时将Tsallis接受准则引入到经典模拟退火当中。其思路是用遗传算法的结果来制约模拟退火算法产生的随机状态,然后由模拟退火的接受准则以及产生的随机状态函数对遗传算法的种群进行更新,从而找到全局近似最优解。实验结果证明,改进算法与单一遗传算法以及经典模拟退火算法相比,其收敛速度和适应度更好,找到全局近似最优解的概率更大。

Abstract:

Hardware/software partitioning is the key issue of Reconfigurable Instruction Sets Processor(RISP) in hardware/software co-design. By comparing with Simulated Annealing Algorithm(SA) and Genetic Algorithm(GA), a hybrid algorithm is proposed , which combines the merits of this two algorithm. Meanwhile, the object function of GA is improved and Tsallis accepting criterion is used in SA. The essence of the algorithm contains two points. On one hand, the random state formed in SA is restricted by the result of GA. On the other hand, the population for GA is updated by the function that formed in SA according to the accepting criterion and random state. Compared to the pure GA and classical SA, the final experimental results indicate that using improved hybrid algorithm can significantly accelerate the convergence speed and increase the ability of getting an approximately optimal solution.