• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2013, Vol. 35 ›› Issue (9): 20-26.

• 论文 • 上一篇    下一篇

MPI非阻塞广播算法及性能研究

严忻恺,郝子宇,吴东,谢向辉   

  1. (数学工程和先进计算国家重点实验室,江苏 无锡 214125)
  • 收稿日期:2013-04-23 修回日期:2013-07-20 出版日期:2013-09-25 发布日期:2013-09-25

Research of MPI nonblocking broadcast
algorithm and its performance analysis     

 YAN Xinkai,HAO Ziyu,WU Dong,XIE Xianghui   

  1. (State Key Laboratory of Mathematical Engineering and Advanced Computing,Wuxi 214125,China)
  • Received:2013-04-23 Revised:2013-07-20 Online:2013-09-25 Published:2013-09-25

摘要:

MPI的3.0版新增了非阻塞集合通信。非阻塞集合通信兼顾非阻塞和集合通信的特点,与阻塞集合通信相比具有更低的同步开销,能够实现更多的计算通信重叠,带来性能提升。以广播为例详细介绍了广播通信的不同算法实现,比较了非阻塞与阻塞广播底层控制管理方法并进行了实验分析,提出了实现改进方法。

关键词: MPI;非阻塞广播;同步开销;重叠

Abstract:

The MPI3.0 standard contains nonblocking collectives. Nonblocking collectives captures the features of both nonblocking communication and collective communication. Compare with blocking collectives, nonblocking collectives have lower synchronization overhead and can realize the overlap between computing and communication in the program so as to improve the performance. The paper describes different algorithm implementations of broadcast, compares the lowlevel management between nonblocking and blocking broadcast, analyzes the experiment results and finally offers some optimizations.

Key words: MPI;nonblocking broadcast, synchronization overhead;overlap