面向峰值电流优化的FSM状态分配

有限状态机(FSM)状态分配与峰值电流有密切关系.针对峰值电流过大易导致电路失效的问题,提出2种优化峰值电流的方法.1)提出一种考虑峰值电流和动态功耗的成本函数,采用遗传算法对两者进行同时优化；2)首先基于遗传算法得到功耗优化后的状态分配解,然后提出基于布尔可满足性(SAT)的启发式算法对功耗优化的状态分配进行重新编码,以降低峰值电流.将这2种方法应用于LGSynth93标准电路的实验结果表明,与传统的功耗优化算法相比,第1种方法虽然功耗略有增加,但能有效地降低峰值电流；第2种方法在实现不增加功耗额外开销的前提下能有效地降低峰值电流,并可将其有效地集成到不同的FSM功耗优化算法中,获得低功耗、低峰值电流的FSM状态分配解.

FSM State Assignment for Peak Current Optimization

State assignment has close relationship with peak current of finite state machines(FSMs).High peak current may result in circuit failure.Against this issue,two peak current optimization methods are proposed based on FSM state assignment.1) A new cost function including switching power and peak current is proposed and a genetic algorithm(GA) based state assignment method is proposed to optimize both power dissipation and peak current.2) First,a power optimized state assignment is obtained by GA,then a SAT-based heuristic algorithm is proposed to re-encode the solution to optimize peak current.The algorithms have been applied to LGSynth93 benchmarks.The experimental results indicated that the first method can reduce peak current efficiently traded with slight power dissipation increase,the second method can reduce peak current significantly with no extra power dissipation overhead.Furthermore,the latter method can be integrated in various FSM power optimization algorithms and implement both low power and low peak current simultaneously.