| In the development of multi-core technology,how to optimize the task scheduling problem and improve the multi-core parallel computing capability has always been the focus of attention.In the traditional task scheduling problem research,traditional methods such as task replication,task clustering,and list scheduling often cannot fully reflect the composition of the solution space,and it is difficult to obtain good scheduling results from possible scheduling solutions at one time.Therefore,the search algorithm for such problems has become the main direction of research in this area.In this paper,the genetic algorithm in the intelligent search algorithm is used as the method for searching the optimal scheduling solution,and two search algorithms GLPGA and CLGA based on the global list and cluster-list hybrid are designed.The GLPGA algorithm uses the island model in the parallel genetic algorithm,uses different genetic operators to search the global scheduling list in parallel,and expands the search range of the better solution space by means of migration;the CLGA algorithm uses a combination of clustering and list scheduling.The optimal assignment method of solving tasks to computing units is used to determine the execution priority of tasks in each cluster through the global task list,and an adaptive convergence judgment flag is used to actively end the search process to find the optimal solution.In the multi-core system model,this paper uses a multi-core model based on the 2D-Mesh structure in the on-chip network communication architecture to conduct behavior-level modeling of the communication process in task processing.The algorithm uses a large number of random static task graphs to simulate the scheduling results under the 2D-mesh structure model.Tests show that the algorithm's on-chip normalized scheduling length is less than 1.9,and the parallelism speedup is up to 3.9.It has good scheduling performance and is superior to Existing scheduling algorithm. |
PDF Full Text Request