Analysis And Research Of Scheduling Algorithm For Asymmetric Multi-core Processors

The Asymmetric Multicore Processors(AMP) is better in improving performance and reducing energy consumption, compared with the Symmetric Multicore Processors(SMP). There has been, up to now, few researches on the scheduling algorithm of APM, and thus the study of efficient utilization of its advantages is of great importance.In this article, the task scheduling algorithms of AMP is studied. The advantages and disadvantages of adaptive scheduling algorithm is analyzed and ways to optimize its performance are proposed. To begin with, the function of average stall time per instruction is optimized. The overall execution time of every task is computed based on the type of the instructions, and thus making the function a better simulation of real operation conditions. Moreover, a new formula for calculating the load of each core in the AMP when balance is achieved is proposed based on the concept of relative load. Finally, the method of thread migration is improved. In the updated method, instead of randomly comparing two cores, in each migration attempt, only the core with the maximum load and the core with the minimum load are considered and threads on them are migrated if the difference between their relative load is greater than 1. As a result, the overload of the current core is avoided, less thread migration attempts are needed to reach balance, and the cost of the algorithm is reduced.