第53回流体力学講演会/第39回航空宇宙数値シミュレーション技術シンポジウム (2021年6月30日-7月2日. 日本航空宇宙学会 : 宇宙航空研究開発機構(JAXA)オンライン会議)

The 53rd Fluid Dynamics Conference / the 39th Aerospace Numerical Simulation Symposium (June 30 - July 2, 2021. The Japan Society for Aeronautical and Space Sciences : Japan Aerospace Exploration Agency (JAXA), Online meeting)

Over the past few years, GPGPU has proven to be a major hardware improvement in high performance computing, and hence to be a way to reach exascale. GPU has found an application in the domain of CFD, where it allows to run larger, more detailed, and more complex simulations at an affordable computational time. However, because GPU programming has its own paradigm, adapting an existing solver with a large source code can be quite challenging. In this aspect, a directive-based language such as OpenACC offers an interesting compromise to obtain good performances without having to alter the solver structure. In this paper, we propose to accelerate the FASTER solver, developed by JAXA, and make it run the NASA-CRM case on a single GPU with a speedup of 23 compared to a single CPU core execution.

形態: カラー図版あり

Physical characteristics: Original contains color illustrations

資料番号: AA2130027026

レポート番号: JAXA-SP-21-008