Floorplan design using a hierarchical neural learning algorithm

Presents a novel floorplanning approach realized by using a neural learning algorithm in a hierarchically organized way. The connection nets of functional modules are described by means of a neural net to obtain a global optimum solution, and the shape and dimensions of each module are simultaneously considered by means of subordinate neural nets to obtain the final floorplan which is a partition for general modules with arbitrary shape. With this hierarchical neural model, not only a fast solution but also a very tight floorplan can be attained. This approach is also suitable for the macrocell placement because of its similarity to the floorplanning. The simulation has been programmed in C and implemented on VAX/VMS environments