Research On Task Scheduling Model For Multi-core Processors

With rapid development of multi-core processors technology,multi-core processors not only performance than traditional single-core processor,and combines the advantage of low power consumption,these advantages make multi-core processors are increasingly used in various fields.For multicore processors,the key to performance is task scheduling.Domestic and foreign scholars have proposed a variety of task scheduling models and algorithms,each with their own merits and demerits,but few have conducted task scheduling research in the case of Shared resource constraints.This paper studies the multi-core task scheduling model with Shared resource constraints.The task scheduling of multi-core processors has been proved to be a NP hard problem,and most of the existing solutions are using heuristic algorithms.In these heuristic algorithms,the saving algorithm is an algorithm with high scheduling success rate.However,the saving algorithm does not consider the parallelism between tasks,but only USES the task deadline to make judgment,and does not fully utilize the constraint of Shared resources.Correlation between tasks is bigger,exclusive resources more cases,access to the exclusive resources between nuclear overhead is very large,easy to make the task for a long time waiting for the exclusive resources,reduce the utilization of nuclear and scheduling success rate decreases.Aiming at this problem,this paper introduced the concept of task correlation,using the task relevant to determine the size of the correlation between tasks,Shared mutexes resources,to the larger task relatedness among multiple tasks scheduling as far as possible to the same nuclear,to reduce the correlation between tasks of exclusive access to overhead.The experiment proves that the improved saving algorithm has a smaller mutual exclusion access cost than the original algorithm in the case of more Shared resources.In the existing task scheduling model,the most common is the centralized scheduling model,which is adopted by the saving algorithm.This model has a centralized global scheduler,which dispatches tasks to each calculation kernel for processing.The domestic and foreign scholars only compare the utilization rate of the calculation kernel with the task scheduling model,and neglect the utilization rate of the scheduling kernel,resulting in the imbalance of processing power between the scheduling kernel and the calculation kernel.This paper proposes a new task scheduling model to solve this problem.While most task scheduling models use the task queues that exist in the Shared data area as the primary means of internuclear task delivery,the performance of the task queue is critical.In existing task queue,MS algorithm is a classical producers over consumers more queue,but MS algorithm in the lack of pseudo Shared cache,in this paper to solve this problem and will improve the MS algorithm is applied to the new task scheduling model is put forward.Finally,the task scheduling model proposed in this paper has a more balanced processing capacity than the original centralized scheduling model.