A probability-based partitioning algorithm, PROP, was introduced in [5] that achieved large improvements over traditional "deterministic" iterative-improvement techniques like FM [7] and LA [10] . While PROP's gain function has a greater futuristic component than FM or LA, it incorporates spatially local information-only information on the removal probabilities of adjacent nets of a cell is used in its gain computation. This prevents a higher-level view of non-local structures. Also, giving

more »

... orm weights to all nets, results in an inability to differentiate between the futuristic benefit of removing one net from another. In this paper, we present a more sophisticated partitioner DEEP-PROP that incorporates more non-local (second-order) structural information than PROP. The second-order information is incorporated into cell gains as well as variable net weights-the latter helps to focus future cell moves in a cluster around the currently moved cell and thus better utilizes the information that led to its selection. A lower-complexity version, VAR-PROP, that also uses dynamically assigned variable net weights, but based on first-order information, has also been developed. Both versions yield significant improvements over PROP on the ACM/SIGDA benchmark suite. DEEP-PROP yields mincut improvements of as much as 39% for large circuits and an average improvement of 20% over all circuits; it is about 3.8 times slower than PROP, which is very fast. VAR-PROP, which has a much lower computational complexity than DEEP-PROP, yields maximum and average mincut improvements over PROP of 27% and 12%, respectively, while being only about 1.14 times slower.