天体物理成团研究中的非规则访存优化.pdf

ISSN 1673-9418 CODEN JKYTA8
Journal of Frontiers of Computer Science and Technology
E-mail:fcst(avip.163.com
http://wwv
1673-9418/2017/11(01)-(0080-11
doi:10.3778j.isSn.1673-9418.1512078
Tel:+86-10-89056056
天体物理成闭研究中的非规则访存优化
郝赫,司雨蒙,韦建文,文敏华,林新华
1.上海交通大学高性能计算中心,上海200240
2. NVIDIA Technology Center Asia Pacific, Singapore 999002
Optimizing Irregular Memory Access in Astrophysical Clustering Studies
HAO He, SI Yumeng, WEI Jianwen, WEN Minhua, LIN Xinhua
1. Center for High Performance Computing, Shanghai Jiao Tong University, Shanghai 200240, China
2. NVIDIA Technology Center Asia Pacific, Singapore 999002
Corresponding author: E-mail: james(@sjtu.edu. cr
HAO He, SI Yumeng, WEI Jianwen, et al. Optimizing irregular memory access in astrophysical clustering
studies. Journal of Frontiers of Computer Science and Technology, 2017, 11(1): 80-90.
Abstract: Halo-based galaxy group finder(HGGF)tries to find galaxies in the same dark matter halo which is not
directly visible. It plays a very important role in the research of large-scale structure of the universe. However, be
cause of the growth of data scale, it s extremely necessary to increase the running speed by optimizing the group
finder coding algorithm. After a thorough investigation on the original HGGF code, it is found that the kernel part of
the algorithm is seriously affected by the irregular memory access. This paper proposes a specific data pre-sorting
approach and analyzes how it affects the process of memory access according to the structure of the algorithm and
the irregular memory access pattern. Moreover, this paper uses data alignment and loop fission to optimize the mem-
ory access as well as improving the efficiency of Openmp with load balance and mutex privatization. Eventually the
HGGF application gets 11.6 times speedup on 12 threads, and gets better weak scalability. The following is the origi-
nal contributions: (1 )Analyze the irregular memory access of the HGGF application (2)Propose and analyze the
data pre-sorting (3 )Improve the parallel performance of HGGF application with another four approaches including
data alignment, loop fission, load balance and mutex privatization
The National High Technology Research and Development Program of China under Grant No.2014A01A302(国家高技术研究发
展计划(863计划); the Project of Japan Society for the Promotion of Science RONPAKU Fellowship(日本学术振兴会项日)
Received 2015-11 Accepted 2016-01
CNKT网络优先出版:2016-01-07,htp:/www.cnki.net/kcms/detail1.602.TP.20160107.1540.00.html