A Statistical Model for Net Length Estimation

The prerouting estimation of net length is very important to the physical design since the estimated length can be employed as a figure-of-merit of the placement process, as an evaluation of the placement result, and in the calculation of capacitance in the predictive timing analysis. The traditional methods, bounded rectangle method, minimal spanning tree method and minimal Steiner tree method, result in either poor estimation or time consuming processing. Moreover, they do not consider the interference between each net and the technology influence. A statistical model for net length estimation has been developed to overcome these deficiencies. This model provides an unbiased estimate with 7 percent relative root-mean-squares error on an average.