Optimization of the power/ground network wire-sizing and spacing based on sequential network simplex algorithm

This paper presents a fast algorithm to optimize both the widths and lengths of power/ground networks under reliability and power dip/ground bounce constraints. The space-sizing which allows the length to change gives more flexibility in solving practical problems. There are two major contributions of this paper. First, we prove that for general topology, a relaxed version of this problem is also convex. Second, we present the sequential network simplex algorithm which can solve those problems with extreme efficiency. Experimental results on several large scale problems, using a PC with a 500-MHZ Pentium III processor, show that our algorithm can solve problems with hundreds of thousands of variables within a few minutes and has a speed improvement of 25+ over sequential linear programming. Experimental results also show that about 50% of the power delivery area can be reduced using our algorithm.