Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Ishizawa, Akihiro*; Idomura, Yasuhiro; Imadera, Kenji*; Kasuya, Naohiro*; Kanno, Ryutaro*; Satake, Shinsuke*; Tatsuno, Tomoya*; Nakata, Motoki*; Nunami, Masanori*; Maeyama, Shinya*; et al.
Purazuma, Kaku Yugo Gakkai-Shi, 92(3), p.157 - 210, 2016/03
The high-performance computer system Helios which is located at The Computational Simulation Centre (CSC) in The International Fusion Energy Research Centre (IFERC) started its operation in January 2012 under the Broader Approach (BA) agreement between Japan and the EU. The Helios system has been used for magnetised fusion related simulation studies in the EU and Japan and has kept high average usage rate. As a result, the Helios system has contributed to many research products in a wide range of research areas from core plasma physics to reactor material and reactor engineering. This project review gives a short catalogue of domestic simulation research projects. First, we outline the IFERC-CSC project. After that, shown are objectives of the research projects, numerical schemes used in simulation codes, obtained results and necessary computations in future.
Nuga, Hideo; Matsuyama, Akinobu; Yagi, Masatoshi; Fukuyama, Atsushi*
Plasma and Fusion Research (Internet), 10, p.1203006_1 - 1203006_2, 2015/01
Fujita, Norihisa*; Nuga, Hideo*; Boku, Taisuke*; Idomura, Yasuhiro
Proceedings of 2014 IEEE 28th International Parallel & Distributed Processing Symposium Workshops (Internet), p.1266 - 1274, 2014/12
Times Cited Count:1 Percentile:30.15GT5D is a nuclear fusion simulation program which aims to analyze the turbulence phenomena in tokamak plasma. In this research, we optimize it for GPU clusters with multiple GPUs on a node. Based on the profile result of GT5D on a CPU node, we decide to offload the whole of the time development part of the program to GPUs except MPI communication. Four NVIDIA M2090 GPUs get 3.35 times faster performance in maximum in function level evaluation, and 1.91 times faster performance in total than two Intel Xeon E5-2670 (SandyBridge) CPU with 8 cores. It includes 63% performance gain with overlapping the communication between MPI processes with GPU calculation.
Fujita, Norihisa*; Nuga, Hideo*; Boku, Taisuke*; Idomura, Yasuhiro
Proceedings of 19th IEEE International Conference on Parallel and Distributed Systems (ICPDCS 2013) (USB Flash Drive), 2 Pages, 2013/12
Takase, Yuichi*; Ide, Shunsuke; Kamada, Yutaka; Kubo, Hirotaka; Mitarai, Osamu*; Nuga, Hideo*; Sakamoto, Yoshiteru; Suzuki, Takahiro; Takenaga, Hidenobu; JT-60 Team
Proceedings of 21st IAEA Fusion Energy Conference (FEC 2006) (CD-ROM), 8 Pages, 2007/03
A self-sustained state driven by the bootstrap current was achieved in JT-60U. Based on a high confinement reversed magnetic shear plasma, a plasma in which the bootstrap current frafciton of almost 100 percent (plasma current 0.55 MA) was successfully maintained for 2.5 s. The dynamic response of a completely self-driven system with negligible external current drive was studied. Repetition of beta collapses and recoveries were often observed. Furthermore, discharges in which the bootstrap current fraction can be expected to be larger than 100 percent were obtained.
Fujita, Norihisa*; Nuga, Hideo*; Idomura, Yasuhiro; Boku, Taisuke*
no journal, ,
In this work, a fusion plasma simulation code GT5D is ported on massively parallel GPU clusters, and its computational performance is evaluated. On the HA-PACS system at Univ. Tsukuba, which has found GPUs per node, major kernels are ported on the GPU memory, and communication between GPUs are overlapped with computations. Although processing performances of the major kernels are improved by 2.3 times on average, an improvement of the total code performance is still limited to 1.3 times, because 70% of kernels are ported at the present. However, the total code performance is slightly improved to 1.85 times by applying the communication overlap technique.
Fujita, Norihisa*; Nuga, Hideo*; Boku, Taisuke*; Idomura, Yasuhiro
no journal, ,
A fusion plasma turbulence simulation code GT5D is optimized for GPU cluster systems with multiple GPUs on a single node. All the calculation loops are ported on GPUs, and communication overlap techniques are applied for main MPI communications. On the HA-PACS system, GPU-CPU computation is faster than CPU-only computation by 2.03 times.
Yagi, Masatoshi; Matsuyama, Akinobu; Wakasa, Arimitsu*; Nuga, Hideo
no journal, ,
Matsuyama, Akinobu; Aiba, Nobuyuki; Nuga, Hideo; Ishii, Yasutomo; Yagi, Masatoshi
no journal, ,
Yagi, Masatoshi; Matsuyama, Akinobu; Nuga, Hideo; Wakasa, Arimitsu*; Fukuyama, Atsushi*
no journal, ,
no abstracts in English
Nuga, Hideo; Matsuyama, Akinobu; Shibata, Yoshihide; Yagi, Masatoshi; Kawano, Yasunori; Fukuyama, Atsushi*
no journal, ,
no abstracts in English
Nuga, Hideo; Matsuyama, Akinobu; Yagi, Masatoshi; Fukuyama, Atsushi*
no journal, ,
no abstracts in English
Yagi, Masatoshi; Matsuyama, Akinobu; Nuga, Hideo; Aiba, Nobuyuki; Ishii, Yasutomo; Fukuyama, Atsushi*
no journal, ,
The current status and issues on disruption research are overviewed in the view point of experiment and DEMO reactor design. Following them, the domestic activity on theory and simulation research aimed for the disruption control will be explained. The progress of simulation research on the runaway electrons and the integrated transport simulation of disruption control are presented. Finally, the urgent issues on the disruption research will be discussed.
Matsuyama, Akinobu; Yagi, Masatoshi; Aiba, Nobuyuki; Nuga, Hideo; Ishii, Yasutomo
no journal, ,
no abstracts in English
Nuga, Hideo; Matsuyama, Akinobu; Yagi, Masatoshi; Fukuyama, Atsushi*
no journal, ,
no abstracts in English
Nuga, Hideo; Matsuyama, Akinobu; Yagi, Masatoshi; Fukuyama, Atsushi*
no journal, ,
no abstracts in English