Circuit placement in arbitrarily-shaped region using self-organization

Effective circuit placement in rectilinear or arbitrarily shaped region is achieved by preplacement of some macro cells. The method is based on the self-organization process proposed by T. Kohonen (Self Organization and Associati Memory, Springer-Verlag, 2nd ed., 1988). The self-organization process is a learning algorithm for neural networks that adjusts the weights of synapses (links) connecting nodes and inputs so that nodes connected closely topologically are sensitive to inputs having similar properties. The authors obtained a good placement result in an arbitrarily shaped region when the cells of the circuit and their positions (x- and y-coordinates) have a one-to-one correspondence with the nodes and the weighted pairs of synapses connected to the nodes, respectively. This method can be easily extended to circuit placement over a nonplanar surface