一种基于模拟退火算法的硬件/软件分解方法

嵌入式系统设计的一个重要环节是硬件/软件分解,通常在系统设计初期解决该问题·硬件/软件分解问题是NP完全问题,无法在有效时间内寻找到满意的分解方案·基于此提出了一个基于模拟退火算法的硬件/软件分解方法,采用了相邻块通讯分解模型以及由单处理器和专用硬件模块构成的目标结构,解决了硬件约束条件下系统执行速度的优化问题·为寻找到性能优良的分解方案,定义了分解优化目标及代价函数,确定了初始温度、终止温度、温度衰减函数等几个控制模拟退火的关键参数,并探讨了模拟退火的停止规则·与贪心算法相比,该方法可以在合理的时间内为嵌入式系统寻找到满意的解决方案·

Method for Hardware/Software Partitioning Using Simulated Annealing Algorithm

One of the key issues for embedded system design is hardware/software partitioning, which should be considered at the beginning of the design. This is a NP complete problem, and could not get good solution generally in reasonable time. Hardware/software partitioning approach based on simulated annealing algorithm was studied. The problem of minimizing system execution time was solved with a hardware constraint. The approach adopts adjacent block communication partitioning model and the target architecture which consists of a single microprocessor and several application specific hardware components. The optimal object and the cost function were defined for partitioning towards the better partitioning alternative. The parameters that control the simulated annealing such as initial temperature, holding time and cooling ratio were examined. Compared with the greedy algorithm, this approach can get good solution for embedded system in timely manner.