The Research Of Task Assignment And Data Allocation In Multi-core Embedded System With Hybrid SPM

With the rise of the mobile Internet,multi-core embedded system gets great progress.At the meanwhile,applications are becoming more and more demanding for the performance and endurance of multi-core embedded system.The research on storage and optimization of Embedded system is becoming more and more important.But the traditional SRAM(Static Random Access Memory)restricts the development of multi-core embedded system because of it's high energy leakage and low integration density.STT-RAM(Spin-transfer Torque Random Access Memory)is a promising nonvolatile memory which has high density and low energy leakage.However,its asymmetric characteristics of read and write reduce its possibility as a successor of SRAM.So the optimization technology based on STT-RAM is highly valuable.In this paper,a STT-RAM based hybrid SPM(Scratchpad memory)is used as on-chip storage.We focus on the performance and energy consumption problem in multi-core embedded system with STT-RAM based hybrid SPM,study task assignment and data allocation methods.Our main work includes:?We propose an optimization method to reduce the energy consumption of multicore embedded system with STT-RAM based hybrid SPM.The method includes an integer non-linear programming(INLP)model and a greedy algorithm,the Minimal Access Energy Cost Data Allocation(MAEDA)algorithm to obtain appropriate data allocation and task assignment solution that effectively reduce data access energy consumption.? We propose a mixed integer quadratic constraint programming(MIQCP)formulation to reduce the schedule length of multi-core embedded system with STTRAM based hybrid SPM.MIQCP formulation finds the optimal solution of task assignment and data allocation and scheduling to enhance the parallelism of memory access.And finally improve the performance of multi-core embedded system.?We propose a heuristic algorithm to reduce memory access time.It divides the optimization of memory access into two sub problems: data allocation and task scheduling.The algorithm obtains sub-optimal data allocation and task assignment and scheduling solution in polynomial time,which reduces schedule length and improves system performance.The simulation results show that our proposed methods can optimize the data allocation and task assignment in multi-core embedded system with STT-RAM based hybrid SPM to improve energy saving and the system performance respectively.