The construction and management of template library for coarse-grained reconfigurable system

Template-based task partitioning and scheduling can dramatically improve the scheduling efficiency and resource utilization in the coarse-grained reconfigurable system. However, when the number of templates is large, the effective organization and management of the templates is a hard problem. Here a template library based method is employed to address the above problem. Various templates are organized and managed within the database; an improved canonical labeling algorithm for DFG (data flow graph) is provided to speed up the template insertion and query by avoiding numerous graph isomorphism operations; an automatic tool is also designed and implemented for the template execution on the RCA (reconfigurable cell array), including the effective methods for context words generation, storage and matching. Experiments are carried out and the results show that the provided method can effectively organize and manage the templates. Moreover, the experiments also indicate that the methods can speed up the template insertion and query operations dramatically.