Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Yamada, Susumu; Imamura, Toshiyuki*; Machida, Masahiko
Proceedings of International Conference for High Performance Computing, Networking, Storage and Analysis (SC '11) (USB Flash Drive), 10 Pages, 2011/11
One of the most fascinating issues in modern condensed matter physics is to understand highly-correlated electronic structures and propose their novel device designs toward the reduced carbon-dioxide future. Among various developed numerical approaches for highly-correlated electrons, the density matrix renormalization group (DMRG) has been widely accepted as the most promising numerical scheme compared to Monte Carlo and exact diagonalization in terms of accuracy and accessible system size. In fact, DMRG almost perfectly resolves one-dimensional chain like long quantum systems. In this paper, we suggest its extended approach toward higher-dimensional systems by high-performance computing techniques. The computing target in DMRG is a huge non-uniform sparse matrix diagonalization. In order to efficiently parallelize the part, we implement communication step doubling together with reuse of the mid-point data between the doubled two steps to avoid severe bottleneck of all-to-all communications essential for the diagonalization. The technique is successful even for clusters composed of more than 1000 cores and offers a trustworthy exploration way for two-dimensional highly-correlated systems.
Idomura, Yasuhiro; Jolliet, S.*
Proceedings of International Conference for High Performance Computing, Networking, Storage and Analysis (SC '11) (USB Flash Drive), 9 Pages, 2011/11
Plasma turbulence simulations based on the five dimensional (5D) gyrokinetic equation, which is the first principle model of fusion plasmas, are standard tools in analyzing turbulent transport phenomena. However, 5D turbulence simulations at the ITER size requires extreme scale computation, and therefore, an efficient use of massively parallel computers has been an important issue in computational fusion science. In GT5D, which is a 5D Eulerian code developed by JAEA, 5D phase space is parallelized using multi-dimensional domain decomposition based on physical symmetry properties of the equation system. By implementing this parallel model on a multilayer network consisting of multiple MPI communicators and a SMP layer, the scalability is highly improved, and the sustained performance of Tflops is achieved on 16384 cores of BX900.
Nakajima, Norihiro; Kawamura, Takuma; Takemiya, Hiroshi; Uzawa, Ken; Idomura, Yasuhiro; Yamada, Susumu
no journal, ,
In the International Conference SC '11 (The Annual Conference for High Performance Computing, Networking, Storage and Analysis), it is exhibited and presented as the achievement of research and development in Center of Computational Science and e-Systems of JAEA, that is, in structural analyses, quake-proof analyses, materials simulations, computer science in nuclear engineering, and applications in nuclear science.
Uzawa, Ken
no journal, ,
CCSE/JAEA promotes the research and development of the three-dimensional vibration simulator for seismic evaluation of a nuclear power plant (NPP) to further improve reliability of seismic safety of NPPs. We started R&D of fluid analysis capability as one of element technologies of the three-dimensional vibration simulator since last year. Especially, we have investigated a two-phase flow simulation code to analyze sloshing under earthquake. In this research, we achieved high parallel efficiency up to approximately 950.2 in parallel calculation using 1024 cores by improving an environment for massive parallel calculation. The technical feasibility for massive sloshing simulation including effect of turbulence is more or less established.
Kawamura, Takuma
no journal, ,
We propose color map adjustment technique which refers the physical value of the simulation result. When the physical values distribute disproportionately such as stress concentration, the traditional color mapping allocates many colors to local region where the physical value changes drastically and plain color is allocated to the other region. The subtle variation is possibly missed in this case. In order to generate color map to visualize the subtle variation, manual operation with try and error is required and the color map automatic adjustment is necessary to reduce the operation time. In order to solve this problem, we tried to apply histogram equalization which is the contrast adjustment technique in image processing area. However this technique changes the physical value. Our technique defines volume density function and calculated its inverse of cumulative function. In the result, our technique successfully adjusted the color map without changing the physical value.
Takemiya, Hiroshi
no journal, ,
Overview of the "Next Generation Seismic Simulation for Large-scale Plants such as Nuclear Facilities" project promoted by CCSE in the HPCI strategic program is described. HPCI strategic program is an national project conducted by MEXT to create innovative results by utilizing a 10 Peta flops class supercomputer "KEI". CCSE has joined one of the strategic fields called "Next Generation Manufacturing" to research and develop the next generation seismic simulation code. In this talk, the aim, current status, and future direction of the research work is introduced.