A Hardware/Software Partitioner Using a Dynamically Determined Granularity

Computer aided hardware/software partitioning is one of the key challenges in hardware/software co-design. While previous ap-proaches have used a fixed granularity, i.e. the size of the parti-tioning objects was fixed, we present a partitioning approach that dynamically determines the partitioning granularity to adapt opti-mization steps to application properties and to intermediate opti-mization results. Experiments with simulated annealing optimiza-tion show a faster convergence and far better adaptability to cost function variations than in previous experiments with fixed granu-larity.