A Dynamic Solution for Efficient MPI Collectiveommunications
The performance of the Message Passing Interface collective communication algorithms is a critical issue widely discussed in both academia and industry. In order to achieve and maintain high performance in the MPI implementations even in regard to random system behavior, the collective operations must be adapted to both the cluster platform and the workload of the user program. In this paper we propose the DYN_Alltoall, a dynamic version of the traditional MPI_Alltoall implementation, which is based on performance predictions derived from P-LogP model. The experiments which were performed on clusters equipped with different interconnect networks, Infiniband and Gigabit Ethernet, produced encouraging results with negligible overhead to find the most appropriate algorithm. In most cases, the dynamic Alltoall largely outperforms the traditional MPI implementations on different platforms.
Hyacinthe Nzigou Mamadou Feng Long Gu Vivien Oddou Takeshi Nanri Kazuaki Murakami
Department of Informatics, Kyushu University Momochihama, Fukuoka, 814-0001 Japan Interdisciplinary Graduate School of Engineering Sciences, Kyushu University Kasuga-koen, Kasuga, 81 ISIT Momochihama, Fukuoka, 814-0001 Japan Research Institute for Information Technology, Kyushu University Hakozaki, Fukuoka, 812-8581 Japan Faculty of Information Science and Electrical Engineering, Kyushu University Motooka, Fukuoka, 819-0
国际会议
三亚
英文
3-7
2009-04-24（万方平台首次上网日期，不代表论文的发表时间）