AN ENERGY-AWARE SCHEDULE STRATEGY FOR CMP SYSTEMS

摘要

This paper focuses the power-performance issues of running task set with interdependence on chip multiprocessor (CMP) systems.First, we propose a tri-dimensional coding based self-adaptive parallel (TCSP) genetic algorithm allocating the task set on processor cores to minimize the execution time.Next, we present a dynamic voltage scaling (DVS) procedure that alters the operating voltage by exploiting the slack time of tasks.The simulation experimental results demonstrate that our two-stage energy-aware schedule strategy can efficiently schedule the tasks on CMP systems while saving the system energy obviously.