Clustering for improved system-level functional partitioning

Abstract

Partitioning of system functionality for implementation among multiple system components, such as among hardware and software components, is becoming an increasingly important topic. Various heuristics can accomplish such partitioning. We demonstrate that clustering can be used to merge pieces of functionality before applying other heuristics, resulting in reduced runtimes with little or no loss in quality, and often with improvements in quality. In addition, we show that clustering, when used for N-way partitioning, fills the gap between fast heuristics and highly-optimizing heuristics.