A graph covering method for template based system partition

Template based system partition method by regularity extraction can be applied to hardware and software architecture generation of complex systems, such as multi-core system and ASIP (Application Specific Instruction-set Processor) containing reconfigurable computing. In the previous work, the search of optimal graph cover, which is the key of this method, was just a greedy way and adopted simple criterions for optimization, thereby sub-optimal and limited. This paper presents a graph covering method by adapted genetic algorithms for template based system partition. Different criterions can be set as evaluation models of various applications, and the optimal covers will always be found by the genetic iterative. Experiments show that the genetic algorithms are effective, even to the time constrained optimization problems including allocation, binding and scheduling.