Simulation Code Optimization for DNS of Turbulence toward Peta-scale Computing

2016 Fiscal Year Final Research Report

• Project/Area Number: 26390130
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Computational science
• Research Institution: Kobe University
• Principal Investigator: YOKOKAWA MITSUO 神戸大学, システム情報学研究科, 教授 (70358307)
• Co-Investigator(Renkei-kenkyūsha): ISHIHARA Takashi 名古屋大学, 大学院工学研究科, 准教授 (10262495)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: Direct Simulation / Parallel Computation / Tubulence / Phase Shift Method / K Computer

Outline of Final Research Achievements

A parallel direct numerical simulation (DNS) code was developed on the K computer for solving the Navier-Stokes equations in a box with periodic boundary conditions for three orthogonal axes. The objective of the code is to simulate behaviors of homogeneous, isotropic incompressible turbulent flow, or canonical turbulent flow, which is the most standard flow among turbulent flows without boundary walls. Pesudo-spectral method is used for discretization in 3-dimensional space and forth-order Runge-Kutta method is used for temporal discretization. Hybrid parallelization with both OpenMP and MPI are adopted in the code. The DNS with grid points of 12288 cubed were carried out on the K computer with sustained performance of 2.2% on the 37,376 compute nodes. Simulation data for about Reynolds number 2300 were obtained.

Free Research Field

大規模並列シミュレーション