@inproceedings{oai:jaxa.repo.nii.ac.jp:00004246,
 author = {坂下, 雅秀 and 橋本, 敦 and 松尾, 裕一 and Sakashita, Masahide and Hashimoto, Atsushi and Matsuo, Yuichi},
 book = {宇宙航空研究開発機構特別資料, JAXA Special Publication: Proceedings of 44th Fluid Dynamics Conference / Aerospace Numerical Simulation Symposium 2012},
 month = {Mar},
 note = {第44回流体力学講演会/航空宇宙数値シミュレーション技術シンポジウム2012 (2012年7月5日-6日. 富山国際会議場大手町フォーラム), 富山市, 富山県, 44th Fluid Dynamics Conference / Aerospace Numerical Simulation Symposium 2012 (July 5-6, 2012. Toyama International Conference Center), Toyama Japan, 10 peta-FLOPS supercomputer "K computer" will begin cooperation in this autumn, and investigation of exa-scale computing already has begun. The other hand, memory throughput will not be enough to supercomputing. So, we consider to using Last-Level-Cache(LLC) on the CPU chip efficiently. Generally, the unstructured CFD solver is parallelized by region decomposition with "Halo". Then, small region which can be computed on LLC is made by decomposing recursively to region. And, with transposing data enough at first, data of another region composed are not required later., 形態: カラー図版あり, Physical characteristics: Original contains color illustrations, 資料番号: AA0061958006, レポート番号: JAXA-SP-12-010},
 pages = {31--35},
 publisher = {宇宙航空研究開発機構(JAXA), Japan Aerospace Exploration Agency (JAXA)},
 title = {非構造格子CFDソルバにおけるCPUキャッシュの有効利用法},
 volume = {JAXA-SP-12-010},
 year = {2013}
}