摘要
采用当前方法对软件模块访存压力进行优化时,优化后的软件模块带宽较高、数据传输延时高,存在有效性差的问题。将机器学习应用在软件模块的访存压力优化过程中,提出基于机器学习的软件模块访存压力优化方法。计算链路的使用率,并将计算结果传送到每条流对应的发送端中,发送端根据接收到的信息对发送速率进行调整,实现拥塞控制。采用多目标规划方法,根据预算值和实际值之间存在的偏差,构建软件模块访存压力优化模型,通过二进制粒子群算法对软件模块访存压力优化模型进行求解,实现软件模块访存压力的优化。仿真结果表明,所提方法的带宽高、数据传输延时小,验证了基于机器学习的软件模块访存压力优化方法的有效性。
When the current method was used to optimize the memory access pressure,the software module has high bandwidth and data transmission delay.Therefore,the machine learning was applied in the optimization process of memory access pressure of software module.In this article,a method of optimizing the access memory pressure of software module based on machine learning was put forward.The usage rate of link was calculated,and then the calculation result was transmitted to the sending end corresponding to each stream.The transmitting end adjusted the transmission rate by the received information,so as to control congestion.The multi-objective programming method was used to construct the optimization model of memory access pressure of software module based on the deviation between budgetary value and actual value.The access pressure optimization model was solved by binary particle swarm optimization algorithm.Finally,the memory access pressure of software module was optimized.Simulation results show that the proposed method has high bandwidth and low data transmission delay.
作者
徐瑞龙
祁云嵩
石琳
XU Rui-long;QI Yun-song;SHI Lin(School of Computer/Jiangsu University of Science and Technology Zhenjiang,Jiangsu 212003,China)
出处
《计算机仿真》
北大核心
2020年第2期212-215,274,共5页
Computer Simulation
关键词
机器学习
软件模块
访存压力
压力优化
拥塞控制
Machine learning
Software module
Memory access pressure
Pressure optimization
Congestion control
