Compact moving particle semi-implicit method for incompressible free-surface flow

Wang, Z.; Matsumoto, Toshinori; Duan, G.*; Matsunaga, Takuya*

Computer Methods in Applied Mechanics and Engineering, 414, p.116168_1 - 116168_49, 2023/09

https://doi.org/10.1016/j.cma.2023.116168

Technical Note: Validation of a material assignment method for a retrospective study of carbon-ion radiotherapy using Monte Carlo simulation

Chang, W.*; Koba, Yusuke*; Furuta, Takuya; Yonai, Shunsuke*; Hashimoto, Shintaro; Matsumoto, Shinnosuke*; Sato, Tatsuhiko

Journal of Radiation Research (Internet), 62(5), p.846 - 855, 2021/09

https://doi.org/10.1093/jrr/rrab028

With the aim of developing a revaluation tool of treatment plan in carbon-ion radiotherapy using Monte Carlo (MC) simulation, we propose two methods; one is dedicated to identify realistic-tissue materials from a CT image with satisfying the well-calibrated relationship between CT numbers and stopping power ratio (SPR) provided by TPS, and the other is to estimate dose to water considering the particle- and energy-dependent SPR between realistic tissue materials and water. We validated these proposed methods by computing depth dose distribution in homogeneous and heterogeneous phantoms composed of human tissue materials and water irradiated by a 400 MeV/u carbon beam with 8 cm SOBP using a MC simulation code PHITS and comparing with results of conventional treatment planning system (TPS). Our result suggested that use of water as a surrogate of real tissue materials, which is adopted in conventional TPS, is inadequate for dose estimation from secondary particles because their production rates cannot be scaled by SPR of the primary particle in water. We therefore concluded that the proposed methods can play important roles in the reevaluation of the treatment plans in carbon-ion radiotherapy.

Implementation and performance evaluation of a communication-avoiding GMRES method for stencil-based code on GPU cluster

Matsumoto, Kazuya*; Idomura, Yasuhiro; Ina, Takuya*; Mayumi, Akie; Yamada, Susumu

Journal of Supercomputing, 75(12), p.8115 - 8146, 2019/12

https://doi.org/10.1007/s11227-019-02983-7

A communication-avoiding generalized minimum residual method (CA-GMRES) is implemented on a hybrid CPU-GPU cluster, targeted for the performance acceleration of iterative linear system solver in the gyrokinetic toroidal five-dimensional Eulerian code GT5D. In addition to the CA-GMRES, we implement and evaluate a modified variant of CA-GMRES (M-CA-GMRES) proposed in our previous study to reduce the amount of floating-point calculations. This study demonstrates that beneficial features of the CA-GMRES are in its minimum number of collective communications and its highly efficient calculations based on dense matrix-matrix operations. The performance evaluation is conducted on the Reedbush-L GPU cluster, which contains four NVIDIA Tesla P100 GPUs per compute node. The evaluation results show that the M-CA-GMRES is 1.09x, 1.22x and 1.50x faster than the CA-GMRES, the generalized conjugate residual method (GCR), and the GMRES, respectively, when 64 GPUs are used.

Application of a communication-avoiding generalized minimal residual method to a gyrokinetic five dimensional Eulerian code on many core platforms

Idomura, Yasuhiro; Ina, Takuya*; Mayumi, Akie; Yamada, Susumu; Matsumoto, Kazuya*; Asahi, Yuichi*; Imamura, Toshiyuki*

Proceedings of 8th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems (ScalA 2017), p.7_1 - 7_8, 2017/11

https://doi.org/10.1145/3148226.3148234

A communication-avoiding generalized minimal residual (CA-GMRES) method is applied to the gyrokinetic toroidal five dimensional Eulerian code GT5D, and its performance is compared against the original code with a generalized conjugate residual (GCR) method on the JAEA ICEX (Haswell), the Plasma Simulator (FX100), and the Oakforest-PACS (KNL). The CA-GMRES method has higher arithmetic intensity than the GCR method, and thus, is suitable for future Exa-scale architectures with limited memory and network bandwidths. In the performance evaluation, it is shown that compared with the GCR solver, its computing kernels are accelerated by , and the cost of data reduction communication is reduced from to of the total cost at 1,280 nodes.

An Approach to DNA crystallization using the thermal reversible process of DNA duplexes

Chatake, Toshiyuki*; Sazaki, Gen*; Kikko, Tatsuhiko*; Fujiwara, Satoru; Ishikawa, Takuya*; Matsumoto, Osamu*; Morimoto, Yukio*

Crystal Growth & Design, 10(3), p.1090 - 1095, 2010/03

https://doi.org/10.1021/cg9007075

We propose a technique for DNA crystallization using the thermal reversible process of DNA: a conversion between a double-stranded DNA (dsDNA) and two single-stranded DNAs (ssDNAs) with temperatures. We investigated the solubility of the crystals of a DNA hexamer d(CGCGCG) and their melting temperature, at which the thermal conversion occurs. The results obtained suggest that the conversion from a dsDNA to ssDNAs results in an increase in solubility. It was shown that using this temperature-controlled technique, high-grade single crystals of the DNA hexamer could be obtained from a small amount of DNA samples. This easy-to-apply technique would be superior to the conventional vapor diffusion technique in that it allows the solubility of DNA crystals to be controlled with no need for expensive setups.

Single-crystal growth and magnetic properties of a new ternary uranium compound UNiAl

Haga, Yoshinori; Matsuda, Tatsuma; Ikeda, Shugo; Galatanu, A.; Matsumoto, Takuya*; Sugimoto, Toyonari*; Tada, Toshiji*; Noguchi, Satoru*; Onuki, Yoshichika

Physica B; Condensed Matter, 359-361, p.1006 - 1008, 2005/04

https://doi.org/10.1016/j.physb.2005.01.379

A new ternary uranium-based intermetallic compound UNiAl has been symthesized. It crystallizes in the unique flat orthorhombic structure. Uranium atoms occupy two crystallographic 4 and 8 sites where local chemical environments are quite similar. The temperature dependence of the magnetic susceptibility demonstrates peculiar magnetic anisotropy; along - and -axes obeys the Curie-Weiss law above 23 K, while along -axis is small and temperature independent. At 23 K, only shows a sharp cusp corresponding to the antiferromagnetic ordering, while remains paramagnetic behavior down to 2 K. These results and crystallographical considerations lead to a conclusion that only uranium atoms at the 8 site order antiferromagnetically at = 23 K, while uranium atoms at the site do not order down to 50 mK.

Development of multiple job execution and visualization system on ITBL system infrastructure software and its utilization for parametric studies in environmental modeling

Suzuki, Yoshio; Matsumoto, Nobuko*; Yamagishi, Nobuhiro*; Higuchi, Kenji; Otani, Takayuki; Nagai, Haruyasu; Terada, Hiroaki; Furuno, Akiko; Chino, Masamichi; Kobayashi, Takuya

Computational Science - ICCS 2003, International Conference Saint Petersburg Russian Federation, Melbourne Australia, June 2-4, 2003 Proceedings, p.120 - 129, 2003/00

Information-Technology Based Laboratory (ITBL) project has been propelled as one of e-Japan priority policy programs. The purposes of the project are to share intellectual resources such as remote computers, programs and data in universities and institutes and to support cooperative studies among researchers, building a virtual research environment, ITBL. Japan Atomic Energy Research Institute (JAERI) has been working on installation and management of hardware and development of infrastructure software and applications. As application software, researches on quantum bioinformatics and environmental sciences are carried out. This paper presents utilization of ITBL system infra-structure software for 'Numerical Environment System' which is developed for environmental studies. More effective job executions and visualization are ex-pected by using Task Mapping Editor (TME) and AVS/ITBL, which are tools developed as infrastructure software.

Development of a Monte Carlo dosimetry system for retrospective study of heavay-ion therapy

Furuta, Takuya; Koba, Yusuke*; Chang, W.*; Hashimoto, Shintaro; Yonai, Shunsuke*; Matsumoto, Shinnosuke*; Sato, Tatsuhiko

no journal, , 

Heavy-ion (carbon-ion) therapy has advantages over conventional radiotherapy such as superior dose concentration and better relative biological effectiveness while the secondary particles produced by nuclear reactions between incident carbon ions and matters induce complexity for risk assessment of secondary cancer. For this assessment, precise transport calculation of secondary particles are required so the Monte Carlo transport calculation is desired. We therefore construct a dosimetry system including PHITS as the engine. In this system, the PHITS input is automatically created from the DICOM data sets recorded in the treatment planning. The transport calculation is simulated by PHITS and dose distribution around the tumor but also out-of-filed is computed. This system will be used as retrospective study in National Institute of Radiological Sciences.

Comparison of physical dose and dose-averaged linear energy transfer between treatment planning system and Monte Carlo simulation in carbon-ion radiotherapy

Ishikawa, Akihisa*; Koba, Yusuke*; Furuta, Takuya; Chang, W.*; Hashimoto, Shintaro; Yonai, Shunsuke*; Matsumoto, Shinnosuke*; Sato, Tatsuhiko

no journal, , 

There found to be a relationship between the dose-averaged linear energy transfer LETd and local tumor control in carbon-ion radiotherapy (CIRT). However, only physical dose and biological dose are registered in the past treatment records of CIRT in QST hospital and LETd can not be deduced directly. There is a method to estimate LETd based on RBE-LETd-fitted function but some problems such as non-singularity at the end point of carbon ions are known. On the other hand, we propose a method to reproduce the CIRT by reconstructing the beam transport geometry based on the treatment planning data and conduct Monte Carlo simulation. The LETd can be also computed directly. We therefore compared LETd obtained by Monte Carlo simulation with estimated LETd using the treatment planning data. We found that underestimation around the end point of carbon ions but the influence was local and thus the LETd estimates are valid for the purpose computing in organ scale.

Implementation and evaluation of a communication avoiding Krylov subspace method, CA-GMRES, on HA-PACS/TCA

Matsumoto, Kazuya*; Idomura, Yasuhiro; Ina, Takuya*; Mayumi, Akie; Yamada, Susumu

no journal, , 

Communication avoiding (CA) Krylov methods are promising solutions for communication bottlenecks on supercomputers based on many core processors or accelerators. In this work, we implemented the CA-GMRES method on a GPU cluster, the HA-PACS, and evaluated its performance on a non-symmetric matrix solver from a nuclear CFD code. The result shows that the CA-GMRES method is significantly faster than the conventional Krylov methods such as the GMRES method and the GCR method.