GA-Based Energy-Efficient Real-Time Task Mapping on Multi-Core Processors

In the VFI-based multi-core system, processing cores are partitioned into VFIs (voltage/frequency islands), so that the system is able to realize fine-grained power optimization by using a unique voltage/frequency for each VFI. In this paper, an energy-efficient mapping algorithm called GA-Mapping is presented, which is based on genetic algorithm, to map real-time tasks onto the processing cores in the VFI-based multi-core system. The proposed mapping algorithm converts the task-core mapping into digit string. Moreover, during the genetic process, the energy consumption is progressively optimized through the proposed selection, crossover and mutation operations. Experimental results demonstrate that compared with the existing energy-efficient mapping algorithms, the proposed algorithm has better performance with respect to energy efficiency as well as schedulability.