Research and optimization of task schedulingalgorithm based on heterogeneous multi-coreprocessor

Abstract

Heterogeneous multi-core processor has the ability to switch between differenttypes of cores to perform tasks, which provides more space and possibility forrealizing efficient operation of computer system and improving computer computing power. Current research focuses on heterogeneous multiprocessor systemswith high performance or low power consumption to reduce system energy consumption. However, some studies have shown that excessive voltage reductionmay lead to an increase in transient failure rates, reducing system reliability. Inorder to solve this problem, this paper studies the energy optimal schedulingproblem of HMSS with DVFS under the constraints of minimum time and reliability, and proposes an improved wild horse optimization algorithm (OIWHO),which improves the efficiency of heterogeneous task scheduling and shortens thetask completion time. The algorithm uses the learning and chaos perturbationstrategies based on opposition and crossover strategies to balance the search andutilization capabilities, and can further improve the performance of OIWHO.Compared with previous work, our proposed algorithm has more advantagesthan existing algorithms. Experimental results show that the average computingtime of OIWHO algorithm is 13.38%, 10.90%, 6.97%, 2.39% and 3.21% fasterthan QHA, PSO, GWO, LFD and OIWOA, respectively. Especially when solvinglarge-scale problems, our algorithm takes less time than other algorithms.