Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Center for Computational Science & e-Systems
JAEA-Evaluation 2023-001, 38 Pages, 2023/07
Research on advanced computational science for nuclear applications, based on "the plan to achieve the medium- and long-term goal of the Japan Atomic Energy Agency", has been performed by Center for Computational Science & e-Systems (CCSE), Japan Atomic Energy Agency. CCSE established a committee consisting of external experts and authorities which evaluates and advises toward the future research and development. This report summarizes the results of the R&D performed by CCSE in FY2022 (April 1st, 2022 - March 31st, 2023) and their evaluation by the committee.
Asahi, Yuichi; Padioleau, T.*; Latu, G.*; Bigot, J.*; Grandgirard, V.*; Obrejan, K.*
Dai-36-Kai Suchi Ryutai Rikigaku Shimpojiumu Koen Rombunshu (Internet), 8 Pages, 2022/12
We implement a kinetic plasma simulation code with multiple performance portable frameworks and evaluated its performance on Intel Icelake, NVIDIA V100 and A100 GPUs, and AMD MI100 GPU. Relying on the language standard parallelism stdpar and proposed language standard multi-dimensional array support mdspan, we demonstrate a performance portable implementation without harming the readability and productivity. With stdpar, we obtain a good overall performance for a kinetic plasma mini-application in the range of 20% to the Kokkos version on Icelake, V100, A100 and MI100. We conclude that stdpar can be a good candidate to develop a performance portable and productive code targeting Exascale era platforms, assuming this programming model will be available on AMD and/or Intel GPUs in the future.
Center for Computational Science & e-Systems
JAEA-Evaluation 2022-004, 38 Pages, 2022/11
Research on advanced computational science for nuclear applications, based on "the plan to achieve the mid- and long-term goal of the Japan Atomic Energy Agency", has been performed by Center for Computational Science & e-Systems (CCSE), Japan Atomic Energy Agency. CCSE established a committee consisting of external experts and authorities which evaluates and advises toward the future research and development. This report summarizes the results of the R&D performed by CCSE in FY2021 (April 1st, 2021 - March 31st, 2022) and their evaluation by the committee.
Center for Computational Science & e-Systems
JAEA-Evaluation 2022-003, 61 Pages, 2022/11
Japan Atomic Energy Agency (hereinafter referred to as "JAEA") consults an assessment committee, "Evaluation Committee of Research Activities for Computational Science and Technology Research" (hereinafter referred to as "Committee") for result and in-advance evaluation of "Computational Science and Technology Research", in accordance with "General Guideline for the Evaluation of Government Research and Development (R&D) Activities" by Cabinet Office, Government of Japan, "Guideline for Evaluation of R&D in Ministry of Education, Culture, Sports, Science and Technology" and "Regulation on Conduct for Evaluation of R&D Activities" by the JAEA. In response to the JAEA's request, the Committee assessed the research program of the Center for Computational Science and e-Systems (hereinafter referred to as "CCSE"). The Committee evaluated the management and research activities of the CCSE based on explanatory documents prepared by the CCSE, and oral presentations with questions-and answers.
Asahi, Yuichi; Padioleau, T.*; Latu, G.*; Bigot, J.*; Grandgirard, V.*; Obrejan, K.*
Proceedings of 2022 International Workshop on Performance, Portability, and Productivity in HPC (P3HPC) (Internet), p.68 - 80, 2022/11
Times Cited Count:1 Percentile:0(Computer Science, Theory & Methods)This paper presents the performance portable implementation of a kinetic plasma simulation code with C++ parallel algorithm to run across multiple CPUs and GPUs. Relying on the language standard parallelism stdpar and proposed language standard multi-dimensional array support mdspan, we demonstrate that a performance portable implementation is possible without harming the readability and productivity. We obtain a good overall performance for a mini-application in the range of 20% to the Kokkos version on Intel Icelake, NVIDIA V100, and A100 GPUs. Our conclusion is that stdpar can be a good candidate to develop a performance portable and productive code targeting the Exascale era platform, assuming this approach will be available on AMD and/or Intel GPUs in the future.
Kubo, Kotaro; Fujiwara, Keita*; Tanaka, Yoichi; Hakuta, Yuto*; Arake, Daisuke*; Uchiyama, Tomoaki*; Muramatsu, Ken*
Proceedings of 29th International Conference on Nuclear Engineering (ICONE 29) (Internet), 8 Pages, 2022/08
After the Fukushima Daiichi Nuclear Power Plant accident, the importance of conducting probabilistic risk assessments (PRAs) of external events, especially seismic activities and tsunamis, was recognized. The Japan Atomic Energy Agency has been developing a computational methodology for seismic PRA, called the direct quantification of fault tree using Monte Carlo simulation (DQFM). When appropriate correlation matrices are available for seismic responses and capacities of components, the DQFM makes it possible to consider the effect of correlated failures of components connected through AND and/or OR gates in fault trees, which is practically difficult when methods using analytical solutions or multidimensional numerical integrations are used to obtain minimal cut set probabilities. The usefulness of DQFM has already been demonstrated. Nevertheless, a reduction of the computational time of DQFM would allow the large number of analyses required in PRAs conducted by regulators and/or operators. We; therefore, performed scoping calculations using three different approaches, namely quasi-Monte Carlo sampling, importance sampling, and parallel computing, to improve calculation efficiency. Quasi-Monte Carlo sampling, importance sampling, and parallel computing were applied when calculating the conditional core damage probability of a simplified PRA model of a pressurized water reactor, using the DQFM method. The results indicated that the quasi-Monte Carlo sampling works well at assumed medium and high ground motion levels, importance sampling is suitable for assumed low ground motion level, and that parallel computing enables practical uncertainty and importance analysis. The combined implementation of these improvements in a PRA code is expected to provide a significant acceleration of computation and offers the prospect of practical use of DQFM in risk-informed decision-making.
Center for Computational Science & e-Systems
JAEA-Evaluation 2021-001, 66 Pages, 2021/11
Research on advanced computational science for nuclear applications, based on "the plan to achieve the mid- and long-term goal of the Japan Atomic Energy Agency", has been performed by Center for Computational Science & e-Systems (CCSE), Japan Atomic Energy Agency. CCSE established a committee consisting of external experts and authorities which does research evaluation and advice for the assistance of the future research and development. This report summarizes the results of the R&D performed by CCSE in FY2020 (April 1st, 2020 - March 31st, 2021), the results expected at the end of the 3rd mid and long-term goal period, and the evaluation by the committee on them.
Center for Computational Science & e-Systems
JAEA-Evaluation 2020-002, 37 Pages, 2020/12
Research on advanced computational science for nuclear applications, based on "the plan to achieve the mid and long term goal of the Japan Atomic Energy Agency", has been performed at Center for Computational Science & e-Systems (CCSE), Japan Atomic Energy Agency. CCSE established a committee consisting of outside experts and authorities which does research evaluation and advice for the assistance of the future research and development. This report summarizes the results of the R&D performed at CCSE in FY2019 (April 1st, 2019 - March 31st, 2020) and the evaluation by the committee on them.
Nagaya, Yasunobu; Adachi, Masaaki*
Proceedings of International Conference on Mathematics & Computational Methods Applied to Nuclear Science & Engineering (M&C 2017) (USB Flash Drive), 6 Pages, 2017/04
MVP is a general-purpose Monte Carlo code for neutron and photon transport calculations based on the continuous-energy method. To speed up the MVP code, hybrid parallelization is applied with a message passing interface library MPI and a shared-memory multiprocessing library OpenMP. The performance test has been done for an eigenvalue calculation of a fast reactor subassembly, a fixed-source calculation of a neutron/photon coupled problem and a PWR full core model. Comparisons has been made for MPI only with 4 processes and hybrid parallelism with 4 processes 3 threads. As a result, the hybrid parallelism yields the reduction of elapsed time by 16% to 34% and the used memories are almost the same.
Sasaki, Takayuki*; Ueda, Kenyo*; Saito, Takumi; Aoyagi, Noboru; Kobayashi, Taishi*; Takagi, Ikuji*; Kimura, Takaumi; Tachi, Yukio
Journal of Nuclear Science and Technology, 53(4), p.592 - 601, 2016/04
Times Cited Count:12 Percentile:74.37(Nuclear Science & Technology)The influences of pH and the concentrations of Eu and NaNO on the sorption of Eu to Na-montmorillonite were investigated through batch sorption measurements and time-resolved laser fluorescence spectroscopy (TRLFS). The pH had a little effect on the distribution coefficients (Kd) in 0.01 M NaNO, whereas the Kd strongly depended on pH at 1 M NaNO. A cation exchange model combined with a one-site non-electrostatic surface complexation model was successfully applied to the measured Kd. The TRLFS spectra of Eu sorbed were processed by parallel factor analysis (PARAFAC), which corresponded to one outer-sphere (factor A) and two inner-sphere (factor B and C) complexes. It turned out that factors A and B correspond to Eu sorbed by ion exchange sites and inner-sphere complexation with hydroxyl groups of the edge faces, respectively. Factor C became dominant at relatively high pH and ionic strength and likely correspond to the precipitation of Eu(OH) on the surface.
Furuta, Takuya; Sato, Tatsuhiko; Ogawa, Tatsuhiko; Niita, Koji*; Ishikawa, Kenichi*; Noda, Shigeho*; Takagi, Shu*; Maeyama, Takuya*; Fukunishi, Nobuhisa*; Fukasaku, Kazuaki*; et al.
Proceedings of Joint International Conference on Mathematics and Computation, Supercomputing in Nuclear Applications and the Monte Carlo Method (M&C + SNA + MC 2015) (CD-ROM), 9 Pages, 2015/04
In Particle and Heavy Ion Transport code System PHITS, two parallel computing functions are prepared to reduce the computational time. One is the distributed-memory parallelization using message passing interface (MPI) and the other is the shared-memory parallelization using OpenMP directives. Each function has advantages and disadvantages, and thus, by adopting both functions in PHITS, it is possible to conduct parallel computing suited for needs of users. It is also possible to conduct the hybrid parallelization by the intra-node OpenMP parallelization and the inter-node MPI parallelization in supercomputer systems. Each parallelization functions were explained together with some application results obtained using a workstation and a supercomputer system, K computer at RIKEN.
Furuta, Takuya; Sato, Tatsuhiko
Igaku Butsuri, 35(3), p.264 - 268, 2015/00
Time-consuming Monte Carlo dose calculation becomes feasible owing to the development of computer technology. However, the recent development is due to emergence of the multi-core high performance computers. Therefore, parallel computing becomes a key to achieve good performance of software programs. A Monte Carlo simulation code PHITS contains two parallel computing functions, the distributed-memory parallelization using protocols of MPI and the shared-memory parallelization using OpenMP directives. Users can choose the two functions according to their needs. This paper gives the explanation of the two functions with their advantages and disadvantages. Some test applications are also provided to show their performance using a typical multi-core high performance workstation.
Matsumoto, Taro; Naito, Hiroshi*; Tokuda, Shinji; Kishimoto, Yasuaki
Physics of Plasmas, 12(9), p.092505_1 - 092505_7, 2005/09
Times Cited Count:2 Percentile:6.85(Physics, Fluids & Plasmas)The behavior of the collisionless magnetohydrodynamics modes is investigated by the gyro-kinetic particle simulation in a cylindrical tokamak plasma in the parameter region where the effects of electron inertia and electron parallel compressibility are competitive for magnetic reconnection. Although the linear growth of the internal kink-tearing mode is dominated by the electron inertia, it is found that the growth rate can be nonlinearly accelerated due to the electron parallel compressibility proportional to the ion sound Larmor radius . It is also found that, as decreasing the electron skin depth , the maximum growth rate before the internal collapse saturates independently of the microscopic scales such as and . The acceleration of growth rate is also observed in the nonlinear phase of the double tearing mode.
Nagaya, Yasunobu; Okumura, Keisuke; Mori, Takamasa; Nakagawa, Masayuki
JAERI 1348, 388 Pages, 2005/06
To realize fast and accurate Monte Carlo simulation of neutron and photon transport problems, two vectorized Monte Carlo codes MVP and GMVP have been developed at JAERI. MVP is based on the continuous energy model and GMVP is on the multigroup model. Compared with conventional scalar codes, these codes achieve higher computation speed by a factor of 10 or more on vector supercomputers. Both codes have sufficient functions for production use by adopting accurate physics model, geometry description capability and variance reduction techniques. The first version of the codes was released in 1994. They have been extensively improved and new functions have been implemented. The major improvements and new functions are (1) capability to treat the scattering model expressed with File 6 of the ENDF-6 format, (2) time-dependent tallies, (3) reaction rate calculation with the pointwise response function, (4) flexible source specification, etc. This report describes the physical model, geometry description method used in the codes, new functions and how to use them.
Ashikagaya, Yoshinobu; Kawasaki, Tomokatsu; Yoshino, Toshiaki; Ishida, Keiichi
JAERI-Tech 2005-010, 81 Pages, 2005/03
no abstracts in English
Nakajima, Norihiro; Ono, Nobuaki*; Suzuki, Yoshio*; Kureta, Masatoshi*
Denki Gakkai Rombunshi, C, 124(10), p.2197 - 2198, 2004/10
Our research interest is to implement volume rendering on CAVE system at enough frame rate. It can be implemented on CAVE easily using texture mapping but there are some defects; for example frame rate declines when the view point is close to the data and images projected to the screens become big. Thus we try to find another way to implement it on CAVE. Considering that Onyx300 is a parallel computer and there are some algorisms of volume rendering which improve the frame rate, we make the program that draws stereo images through two of the algorisms and by parallel computing, and displays only two images on CAVE. And we confirm that this program works well and draws stereo images at sufficient frame rates.
Azumi, Masafumi
Purazuma, Kaku Yugo Gakkai-Shi, 80(5), p.378 - 381, 2004/05
Progress of large scale scientific simulation environment in JAERI is briefly described. The expansion of fusion simulation science have been played a key role in the increasing performances of super computers and computer network system in JAERI. Both scalar parallel and vector parallel computer systems are now working in Naka and Tokai sites respectively and, particle and fluid simulation codes developed under the fusion simulation project, NEXT, are running on each system. The storage grid system has been also successfully developed for the effective visualization analysis by remote users. Fusion research is going to enter the new phase of ITER, and the need for the super computer system with higher performance are increasing more than as ever along with the development of reliable simulation models.
Ueshima, Yutaka
JAERI-Conf 2004-003, 62 Pages, 2004/03
no abstracts in English
Advanced Photon Research Center; Center for Promotion of Computational Science and Engineering
JAERI-Conf 2004-002, 81 Pages, 2004/03
no abstracts in English
Nagai, Haruyasu
JAERI-Data/Code 2003-021, 36 Pages, 2003/12
A new method to couple atmosphere and land-surface models using the massage passing interface (MPI) was proposed to develop a atmosphere-land model for studies on heat, water, and material exchanges at around the land surface. A non-hydrostatic atmospheric dynamic model of Pennsylvania State University and National Center for Atmospheric Research (PUS/NCAR-MM5) and a detailed land surface model (SOLVEG) including the surface-layer atmosphere, soil, and vegetation developed at Japan Atomic Energy Research Institute (JAERI) are used as the atmosphere and land-surface models, respectively. Concerning the MPI, a message passing library named Stampi developed at JAERI that can be used between different parallel computers is used. The models are coupled by exchanging calculation results by using MPI on their independent parallel calculations.