A GRASP based algorithm for multi-constraint circuit partitioning

We study the circuit partitioning problem with area and delay constraints to optimize the circuit performance. We integrate delay constraints into the greedy randomized adaptive search procedure (GRASP), so that GRASP can be used to deal with delay problems. The GRASP with area and delay constraints can generate better initial partitions. Then the initial partition is iteratively improved by an extended FM algorithm. Finally we obtain the partition satisfying area and delay constraints. Our hybrid algorithm can solve bi-partition with area and delay constraints successfully. The experimental results indicate that, compared with FM algorithm and clustering algorithm, our hybrid algorithm improves the solution quality in a reasonable computational time, and obtains a better partition, especially in terms of the minimum cut-size and the average cut-size.