异构系统结构力学计算GPU加速性能分析与应用
发布时间：2018-10-11 12:39
【摘要】：高性能计算领域利用GPU加速计算已逐渐发展成为主流应用的普遍功能之一。主流结构力学应用ABAQUS支持GPU通用计算,充分发挥GPU的高性能浮点运算能力与访存带宽,提高软件求解效率。介绍了ABAQUS软件对GPU加速应用的发展历程,在上海超级计算中心三种不同的异构环境下,针对实际算例的GPU加速性能进行了测试,建立计算效率分析方法,分析GPU加速对求解时间、系统资源调用、软件并行效率的影响,并针对采用隐式方法求解百万量级自由度问题时资源的合理使用提出建议。
[Abstract]:In the field of high performance computing, accelerated computing using GPU has gradually developed into one of the universal functions in mainstream applications. ABAQUS is used in mainstream structural mechanics to support GPU universal computing, which makes full use of the high performance floating-point computing capability and memory access bandwidth of GPU, and improves the efficiency of software solution. This paper introduces the development of the application of ABAQUS software to the acceleration of GPU. In three different heterogeneous environments of Shanghai Supercomputing Center, the acceleration performance of GPU is tested for practical examples, and the calculation efficiency analysis method is established. This paper analyzes the effects of GPU acceleration on solving time, system resource transfer and software parallel efficiency, and puts forward some suggestions for the rational use of resources when solving the problem of millions of degrees of freedom by implicit method.
【作者单位】： 上海超级计算中心高性能计算应用技术部;
【基金】：国家863计划项目(2012AA01A308)
【分类号】：TP332;TP311.5
[Abstract]:In the field of high performance computing, accelerated computing using GPU has gradually developed into one of the universal functions in mainstream applications. ABAQUS is used in mainstream structural mechanics to support GPU universal computing, which makes full use of the high performance floating-point computing capability and memory access bandwidth of GPU, and improves the efficiency of software solution. This paper introduces the development of the application of ABAQUS software to the acceleration of GPU. In three different heterogeneous environments of Shanghai Supercomputing Center, the acceleration performance of GPU is tested for practical examples, and the calculation efficiency analysis method is established. This paper analyzes the effects of GPU acceleration on solving time, system resource transfer and software parallel efficiency, and puts forward some suggestions for the rational use of resources when solving the problem of millions of degrees of freedom by implicit method.
【作者单位】： 上海超级计算中心高性能计算应用技术部;
【基金】：国家863计划项目(2012AA01A308)
【分类号】：TP332;TP311.5
【相似文献】
相关期刊论文 前10条
1 辛年;;如何选用386计算机[J];微计算机信息;1988年03期
2 翟晓宇;;论中等师范类院校计算机教学[J];考试(教研版);2009年02期
3 伊扬;;首次采用POWER4+ IBM IntelliStation POWER 275性能跃升3倍[J];每周电脑报;2003年23期
4 晓风;;CAE应用有待深化提高 市场发展前景无限[J];机械工业信息与网络;2008年03期
5 陈小军;张t，
本文编号：2264228
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2264228.html
