加权主元分析在统计互连寄生参数提取中的应用-WenjianYU-清华.PDF

22 11 Vol. 22 No. 11 2010 11 Journal of ComputerAided Design Computer Graphics Nov. 2010 1, 2) 1, 2) * 3) 胡 超 喻文健 Sheldon X D T an , , 1) ( 100084) 2) ( ( ) 100084) 3) ( Department of Electrical Engineering, University of California, Riverside, CA 92521 U SA) ( yuwj@ tsinghua. edu. cn) : , H erm ite , , . , , . , , , 8 CPU 6.7 . : ; H ermite ; ; : T N391.9 Weighted Principal Factor Analysi s in Statistical Interconnect Parasitic Extracti on Hu Chao1, 2) , Yu Wenjian1, 2) * , and Sheldon X D Tan3) 1) (Dep artment of Comp uter Sc ience and Technolog y , Tsinghua Univ ersity , B eij ing 100084) 2) ( Tsingh ua National L a orator y f or I nf or mation S cience and T echnolog y ( TN list) , B eij ing 100084) 3) (Dep artment of E lectr ical Engineering , Univ ersity of Calif ornia, R iv er side, CA 92521 USA ) Abstract: To cope with the problems of statistical parasitic extraction induced by random process variations, the technique of weighted principal factor analysis (wPFA) based on Hermite Polynomial Collocation method is proposed to reduce the number of random variables and improve the computational efficiency . T he parallel computing technique is also applied to further reduce the computational time. Numerical results show that, the wPFA is able to accelerate the statistical extraction using a normal principal factor analysis by several or several tens times. While, the parallel computing experiment on a machine with 8 CPU achieves a speedup of 6.7. Key words: variationaware parasitic extraction; Hermite polynomial collocation method; weighted principal factor analysis; parallel co