Finding best cones from random clusters for FPGA package partitioning

Abstract

The implementation of circuits with FPGAs commonly requires partitioning into several FPGA packages. A good FPGA package partitioning tool minimizes production cost by minimizing packages required without significantly increasing the circuit operational frequency. Prior work has shown that good results can be obtained by partitioning with cone structures. Cone structure partitioning is combined with random cluster creation for better FPGA package partitioning. Experimental results show that the final partitioning algorithm presented, produces better circuit partitions than the high quality Min-Cut tool of MCNC.