期刊文献+

OpenMP程序性能退化的诊断与处理 被引量:1

Performance Degradation Diagnosis and Solution for OpenMP Programs
下载PDF
导出
摘要 为了解决OpenMP程序性能退化问题,本文提出性能退化区和性能退化强度的概念.使用性能退化强度能够剔除非性能退化区并突出执行时间较长的性能退化代码段;同时,性能退化区的分解能够逐步缩小性能退化区并最终准确定位引发性能退化的代码段.去除引发性能退化的根源就能有效改进OpenMP程序的执行性能.实例分析证实了本文提出的OpenMP程序性能退化诊断与处理方法的有效性. In order to solve performance degradation in OpenMP programs, this paper proposes the conception about performance degradation region and performance degradation strength. Using performance degradation strength is able to eliminate code regions whose performance doesn't degrade and give prominence to performance degradation regions that need relatively long execution time. At the same time, the decomposition of performance degradation regions can reduce the scope of performance degradation regions and finally locate the code sections that result in performance degradation. Then ,the performance of OpenMP programs will he improved through removing roots of performance degradation. The analysis of examples proves that the method of performance degradation diagnosis and solution for OpenMP programs proposed in this paper is efficient.
出处 《小型微型计算机系统》 CSCD 北大核心 2005年第9期1664-1668,共5页 Journal of Chinese Computer Systems
基金 国家自然科学基金(69933020)资助 Intel公司OpenMPforORC课题资助.
关键词 OPENMP 性能 退化 分解 OpenMP performance degradation decomposition
  • 相关文献

参考文献9

  • 1OpenMP 2. 0 standard. OpenMP API V2.0[EB/OL]. http://www. openmp, org/.
  • 2William Gropp. A user's view of OpenMP.. The good,the bad,and the ugly[C]. In: Workshop on OpenMP Applications and Tools, San Diego Supercomputer Center, San Diego, California,2000.
  • 3Michael Voss, Rudolf Eigenmann. Reducing parallel overheads through dynamic serialization[C]. In.. 13th International Parallel Processing Symposium and 10th Symposium on Parallel and Distributed Processing, San Juan, Puerto Rico, 1999 : 88-92.
  • 4Marc Gonzeilez,Albert Serra,Xavier Martorell et al. Applying interposition techniques for performance analysis of OpenMP parallel applications [C]. In: Proceedings of the 14th International Parallel & Distributed Processing Symposium (IPDPS'00) ,Cancun,Mexico,2000: 235-240.
  • 5KAI of Intel. GuideView performance analyzer [EB/OL].http://www. kai. com/parallel/kappro/guideview/.
  • 6Malony A, Shende S. Performance technology for complex parallel and distributed systems [C]. In.. Proc. 3rd Workshop on Distributed and Parallel Systems, DAPSYS 2000,"Distributed and Parallel Systems.. From Concepts to Applications," (Eds.G. Kotsis and P. Kacsuk), 37-46,2000.
  • 7Bane M, Riley G. Overheads profiler for OpenMP codes[C].In.. European Workshop on OpenMP (EWOMP 2000) ,September, 2000.
  • 8Intel Corporation. VTune performance environment[EB/OL].http ://www. intel, com/soft ware/product s/vt une/.
  • 9Jin H, Frumkin M, Yan J. The OpenMP implementation of NAS parallel benchmarks and its performance[D]. NAS Technical Report NAS-99-011,October 1999.

同被引文献5

  • 1王家翀,许川,姚栋.蒙特卡罗算法并行计算研究[J].核动力工程,2007,28(4):20-24. 被引量:5
  • 2陈国良.并行计算:结构、算法、编程[M].北京:高等教育出版社.2003.
  • 3NVIDIA CUDA C Programming Guide, Version 3. 2, NVIDIA Corporation,2010.10.
  • 4Thomas H Corrnen,等.算法导论(第二版,影印版)[M].北京:高等教育出版社,2002.5.
  • 5OpenMP API周户指南, Sun Microsystems, Inc. 2005. 11.

引证文献1

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部