Research On The Task Adjustment Algorithm Of Network-on-chip Based On NBTI Aging Effect Perception

Task scheduling is an important problem in the design of network-on-chip(NoC).Negative Bias Temperature Instability(NBTI)can cause an aging effect on transistor devices,which may result in the degraded frequency of processing core as well as the decreased reliability of NoC system.Therefore,the research on NBTI aging-aware task scheduling algorithms for NoC system is a challenging problem.Aiming at the problem of operating frequency degradation and system reliability decrease caused by NBTI aging effect,this thesis proposes a series of NBTI aging-aware task scheduling algorithms,which were used for scheduling tasks to improve application's makespan and system reliability,respectively.First,we derive the frequency degradation according to the variation of device parameters caused by NBTI aging effect,and establish a NBTI aging prediction model to evaluate the task execution time on an aged core.Considering the frequency degradation,an aging-aware scheduling optimization model was constructed with application's makespan as the scheduling objective,and a scheduling algorithm based on particle swarm optimization(PSO)was developed to effectively solve the above optimization problems.In order to mitigate the degradation of system reliability caused by aging effect,a new aging effect-aware scheduling model was constructed to optimize system reliability under a deadline constraint.Then a binary particle swarm optimization(BPSO)mechanism is designed to represent the mapping between a scheduling solution and a particle,for the purpose of improving the efficiency of the scheduling algorithm.Experimental results show that the proposed aging-aware scheduling algorithm for improving makespan can achieve shorter makespan and higher throughput,and the proposed aging-aware scheduling algorithm for improving reliability can achieve higher system reliability and higher throughput.Moreover,the HNOCS(Heterogeneous Network on Chips simulator)simulator is rewritten,and a series of simulation experiments were performed to verify the performance and effectiveness of the proposed algorithm.The aging-aware scheduling algorithms proposed in this paper can effectively improve the efficiency of task execution and enhance the system reliability,and therefore has important practical value and promising application prospects n for the research of reliable scheduling.