实验分析了电源/地线(P/G)网络的随机行走算法,与传统的预优共轭梯度法比较发现,当随机行走求解较少节点时性能较好,但在大规模PiG网络瞬态分析中相对较慢.通过修改随机行走过程进行伴随网络的瞬态分析,提出一种快速计算灵敏度的算法.实验结果表明,该算法计算时间较短,与精确结果误差较小.
Experimental study was conducted on random walk algorithm for power/ground (P/G) network. Compared with the traditional PCG approach, random walk is much faster when solving several nodes, but it is slow for transient analysis of huge P/G network. By analyzing the adjoint network using the modified random walk process, a new method is proposed for P/G network sensitivity computing. The experimental results show that the runtime of our proposed method is fast and the relative error is acceptable.