摘要
人工智能运行平台监控的实时性要求相当高,单CPU难以满足实时人工智能运行平台的实际应用要求。为了改善人工智能运行平台的监控效率,设计了基于“CPU+GPU”的人工智能运行平台的实时监控方法。首先分析了当前人工智能运行平台的监控研究进展,然后设计了人工智能运行平台整体流程,采用GPU作为加速部分,配合CPU共同承担实时监控任务,最后进行了仿真对比测试,相对于单CPU,CPU+GPU的平台监控速度更快,可以满足平台实时控制要求,具有更高的实际应用价值。
The real-time monitoring requirements of the artificial intelligence operation platform are quite high.A single CPU can not meet the real-time requirements of the artificial intelligence operation platform.In order to improve the monitoring efficiency of the artificial intelligence operation platform,a real-time monitoring method of the artificial intelligence operation platform based on“CPU+GPU”is designed.Firstly,the monitoring research progress of the current artificial intelligence operation platform is analyzed,then the overall framework of the artificial intelligence operation platform is designed,which uses GPU as the acceleration part and cooperates with CPU to jointly undertake the real-time monitoring task.Finally,the simulation comparison test is carried out.Compared with a single CPU,the platform monitoring speed of CPU+GPU is faster,which can meet the real-time control requirements of the platform and has higher practical application value.
作者
吴文炤
王卫卫
邱镇
郭庆
程琳
WU Wen-zhao;WANG Wei-wei;QIU Zhen;GUO Qing;CHENG Lin(State Grid Information&Telecommunication Group Co.,Ltd.,Beijing 102200,China;Anhui Jiyuan Software Co.,LTD.,Hefei,Anhui 230088,China)
出处
《计算技术与自动化》
2023年第2期86-90,共5页
Computing Technology and Automation
关键词
人工智能
中央处理器
图形处理器
对比测试
实时控制
artificial intelligence
central processing unit
graphics processing unit
comparative test
real-time control
