MogaMap and DynPack: Multi-Objective mapping and packing algorithms for optimization of area, performance and power consumption in FPGAs

摘要

This paper presents a new technology mapper, MogaMap. The mapper is based on a hybrid approach that uses evolutionary algorithm associated with specific heuristics of the problem in order to find better trade-off results among area, performance and power consumption. Besides that, we present DynPack, a multi-objective clustering tool based on net connections and cluster dynamic status. MogaMap and DynPack have been integrated into VTR and, the pos-route analysis shows that, on average, our technique is able to reduce the number of complex block logics by 8%, the critical path delay by 16% and the power consumption by 15% when compared to state-of-the- art.