基于遗传算法的最大开启电流估计

集成电路设计进入深亚微米阶段后，静态功耗成为低功耗设计中的一个瓶颈．电源门控法可以同时有效地降低动态功耗和静态功耗，是一项具有广阔应用前景的技术．电源门控电路的最大电流是由最大开启电流和最大的正常运行电流决定，它是电路设计的一个十分重要的参数，如何对它进行快速准确的估计已经成为一个新的问题．另外，冒险功耗是电路整体功耗中非常重要的组成部分，该文通过研究发现，在电路开启阶段同样存在冒险，同时消耗了大量的能量．文章考虑了组合电路的冒险现象，提出了一种基于遗传算法的最大开启电流的估计方法，对ISCAS85电路的实验结果表明，电源门控电路的开启最大功耗可能比正常情况下的最大功耗还要大．该文的方法具有较小的复杂性，可以仅用随机模拟的2．77％的时间，获得12．90％的最大开启电流值增量。

Maximum Power-up Current Estimation Based on Genetic Algorithm

When scaling down into deep sub micron of VLSI design, the static power becomes a bottleneck of low power design. Power gating is a powerful and applicable solution to reduce both dynamic power and static power. Maximum current is a significant parameter for power gated circuit designs, which is determined by the maximum of all possible power up and normal switching current. How to estimate the maximum current of a power gated circuit quickly and accurately is a new problem. Hazards are found to consume a significant part of total power during runtime. In this paper, we develop an experiment and observe that hazards exit during powering up as well as normal runtime and dissipate a great deal of energy. A new genetic algorithm based method is proposed to estimate the maximum power up current for combinational circuits considering hazard power. Experiments with the ISCAS85 benchmark circuits show power gated circuits may consume larger maximum current in powering up phase than normal runtime phase and our algorithm can obtain 12.90% improvement of maximum power up current within only 2.77% time than random simulation due to its lower time complexity.