Placement Algorithm by Partitioning for Optimum Rectangular Placement

This paper describes methods for placement of DIP packages on a rectangular grid with minimum crossovers and wire length. The initial step is partitioning the components into subsets, one subset for each row, such that the connection between rows are minimized. An iterative procedure, yielding successively improved partitioning, is used for this purpose. Once the row sets are formed, a process for best placement of the DIP in a row is carried-out. The row is assumed to have either a one row channel for routing wires or two row channels. For both cases a placement which minimized crossovers and line length is obtained. The basic method uses simple branch and bound search procedures for placing the packages within a row set. In the placement procedure the objects being placed are wire nets rather than components. Previous work related to this problem is included in references [1-6].