首页 | 本学科首页   官方微博 | 高级检索  
     

OpenCL计算软件栈评估
引用本文:朱浩,周博洋,卢雪山,杜溢墨. OpenCL计算软件栈评估[J]. 计算机工程与科学, 2021, 43(12): 2105-2114. DOI: 10.3969/j.issn.1007-130X.2021.12.003
作者姓名:朱浩  周博洋  卢雪山  杜溢墨
作者单位:(1.军事科学院国防科技创新研究院,北京 100000;2.国防科技大学计算机学院,湖南 长沙 410073;3.空军后勤部,北京100000; 4.31008部队,北京100091)
基金项目:国家自然科学基金(61802416,61972408,61902407)
摘    要:随着智能计算和大数据应用的发展,人们对GPU等加速部件的需求不断增长.计算软件栈比如CUDA、OpenCL软件栈是能充分发挥GPU硬件性能的关键.考虑计算软件栈未来在国产基础软硬件平台(比如飞腾CPU和麒麟操作系统)上的可移植性和适配性,重点研究OpenCL开源计算软件栈.测试分析OpenCL应用在不同平台上的表现,评估应用在不同OpenCL软件栈上(比如Mesa、ROCm等)进行GPU计算的表现,评估软件栈中驱动、内核等对GPU计算的影响,并且整个测试涵盖了编译、数据传输和内核执行等OpenCL计算各个阶段的时间开销.经过测试评估发现,国产平台更迫切也更适合使用GPU进行加速计算,ROCm是比较理想的OpenCL开源软件栈,有较好的性能和稳定性,并且与闭源软件栈相比存在一定的优化空间.

关 键 词:OpenCL  计算软件栈  GPU计算  国产基础软硬件平台  
收稿时间:2020-11-20
修稿时间:2021-03-02

Evaluation of OpenCL computing software stack
ZHU Hao,ZHOU Bo-yang,LU Xue-shan,DU Yi-mo. Evaluation of OpenCL computing software stack[J]. Computer Engineering & Science, 2021, 43(12): 2105-2114. DOI: 10.3969/j.issn.1007-130X.2021.12.003
Authors:ZHU Hao  ZHOU Bo-yang  LU Xue-shan  DU Yi-mo
Affiliation:(1.Defense Innovation Institute,Academy of Military Sciences,Beijing 100000;2.College of Computer Science and Technology,National University of Defense Technology,Changsha 410073;3.PLA Air Force Logistics Department,Beijing 100000;4.Troop 31008,Beijing 100091,China)
Abstract:With the development of intelligent computing and big data applications, the demand for accelerators such as GPU is increasing. Computing software stacks such as CUDA and OpenCL software stacks are the key to making full use of GPU hardware performance. Considering the portability and implementation of software stacks on domestic fundamental OS and hardware platforms (such as Phytium CPU and Kylin OS) in future, this paper focuses on open-source OpenCL software stacks. The performance of OpenCL applications on different platforms is tested and analyzed. The performance of GPU computing on different OpenCL software stacks, such as Mesa, ROCm, etc., is evaluated. The impact of drivers and kernels in the software stack on GPU computing is evaluated. The entire test covers the time overhead of various stages of OpenCL calculations such as compilation, data transmission, and kernel execution. The test and evaluation found that it is more urgent and more suitable to use GPU for accelerated computing on domestic platforms. ROCm is an ideal OpenCL open source software stack with better performance and stability, and can be further optimized compared with close-source software stacks.
Keywords:OpenCL  computing software stack  GPU computing  domestic fundamental software and hardware platform  
本文献已被 万方数据 等数据库收录!
点击此处可从《计算机工程与科学》浏览原始摘要信息
点击此处可从《计算机工程与科学》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号