Communication overlap techniques for improved strong scaling of gyrokinetic Eulerian code beyond 100k cores on the K-computer

京におけるジャイロ運動論的オイラーコードのストロングスケーリングを10万コア以上に向上するための通信オーバーラップ手法

井戸村 泰宏   ; 仲田 資季; 山田 進  ; 町田 昌彦  ; 今村 俊幸*; 渡邉 智彦*; 沼波 政倫*; 井上 晃*; 堤 重信*; 三吉 郁夫*; 志田 直之*

Idomura, Yasuhiro; Nakata, Motoki; Yamada, Susumu; Machida, Masahiko; Imamura, Toshiyuki*; Watanabe, Tomohiko*; Nunami, Masanori*; Inoue, Hikaru*; Tsutsumi, Shigenobu*; Miyoshi, Ikuo*; Shida, Naoyuki*

A plasma turbulence research based on 5D gyrokinetic simulations is one of the most critical and demanding issues in fusion science. To pioneer new physics regimes both in problem sizes and in time scales, an improvement of strong scaling is essential. Overlap of computations and communications is a promising approach to improve strong scaling, but it often fails on practical applications with conventional MPI libraries. In this work, this classical issue is clarified, and resolved by developing communication overlap techniques with mpi_test and communication threads. These techniques dramatically improve the parallel efficiency of a gyrokinetic Eularian code GT5D on the K-computer, and ITER-size simulations are significantly accelerated.