Optimization of placement by candidate sieving

A new algorithm for standard cell placement problem is presented. It is based on an idea that appropriate step by step reduction of allocatable cells for each component yields a good placement result. This algorithm named candidate sieving is applicable for discrete quadratic assignment problem with some constraint. The results obtained by candidate sieving are compared in terms of wire length and computing time with those obtained by simulated annealing which is considered the most popular method today. In many cases, candidate sieving has yielded shorter wire lengths, which are equal or almost equal to those obtained by simulated annealing, but with much less computing time.