Implementing a Fiduccia-Mattheyses Algorithm

Hi,

I'm trying to think of the best way to implement the Fiduccia-Mattheyses (FM) algorithm. The goal is to get the best partition with the least net cuts. Any ideas would be appreciated. So far I have two partitions of clustered nodes. The clustered nodes only account for 30% of the original number of nodes.

I'm thinking I'll need some sort of cut matrix to keep track of the best net cut and the current partitions. The partitions can not be less than 40% of the original number of nodes. Nodes will be somewhere in the number of 10,000-50,000. I'm also unsure if having the clustered nodes in a weighted form would be easier or not. Not entirely sure how to do that.

Thanks.