Research On Task Allocation And Energy For Heterogeneous Multi-core Processors

Multi-core processors is the trend now. Multi-core processors is divided into homogeneous multi-core and heterogeneous multi-core according to the difference among the system's core structure.But the performance of the homogeneous multi-core has certain limitations:when the system reaches its top,the performance will not increase significantly by adding the cores.But the heterogeneous multi-core processors can supply a good solution to this problem.Heterogeneous multi-core can improve the system throughput and save the energy consumption,so the designers generally believe, that the heterogeneous multi-core will become popular. At the same time,the emergence of multi-core processors system brings new challenges for the task marching and scheduling.With different task matching and scheduling will lead to different time consumption and energy consumption on Heterogeneous multi-core processors,using approprivate task matching and scheduling algorithms can save more energy. Now, it is widely recognized that the most promising of the task matching and scheduling is to use heuristic way to match the tasks firstly, and then to use genetic algorithm to schedule them.Now, the research on task marching and scheduling of heterogeneous multi-core is mainly load balance,but the research on energy consumption is very little.This paper mainly concentrates on saving more energy consumption when meeting the condition of real-time.Firstly, we improve the heuristic way to match the tasks,and then we use genetic algorithm to schedule tasks to reslove the energy consumption problem for the first time.At last,we make the genetic algorithm better to improve the algorithm's real-time.The main contribution of this paper is as follows:1.Take more care of the communication cost of the inter-process, when we allocate the task to the process nodes.Afer improving task groups, the energy consumption caused by communication among the nodes is the lowest;2.Use genetic algorithm to resolve the energy problem on heterogeneous multi-core processors;3.We use OpenMP technology to make genetic algorithm threaded,which(im-proved genetic algorithm)is more suitable for running on heterogeneous multi-core processors environment,accelerate the convergence speed and prepare to save more energy consumption; 4.Mutation operator optimization,we can make sure that the task will migrate from the core which need more time to execute all the tasks(scheduling to this core)to the other core when mutation happens;5.It does not always require a minimum energy consumption one when the genetic algorithm output the best individual.We can consider both time consumption and energy consumption,which can ensure improving the algorithm's real-time.