TAKESHI FUKAYA
| Information Initiative Center Supercomputing | Associate Professor |
Last Updated :2025/01/11
■Researcher basic information
Researchmap personal page
J-Global ID
■Career
Career
- Nov. 2022 - Present
Hokkaido University, Information Initiative Center, Associate Professor, Japan - Apr. 2015 - Oct. 2022
Hokkaido University, Information Initiative Center, 助教 - Nov. 2020 - Mar. 2022
科学技術振興機構, さきがけ研究員 - Oct. 2013 - Mar. 2015
理化学研究所 計算科学研究機構, 研究部門 大規模並列数値計算技術研究チーム, 特別研究員 - Apr. 2012 - Sep. 2013
Kobe University, Graduate School of System Informatics, 特命助教
Educational Background
Committee Memberships
- Apr. 2022 - Present
情報処理学会 ハイパフォーマンスコンピューティング研究会, 運営委員, Society - Apr. 2020 - Present
日本応用数理学会 行列・固有値問題の解法とその応用 研究部会, 幹事, Society - Apr. 2019 - Present
日本応用数理学会 行列・固有値問題の解法とその応用 研究部会, 運営委員, Society - Apr. 2017 - Present
HPCI 連携サービス運営・作業部会, 会員, Others - Apr. 2019 - Mar. 2023
自動チューニング研究会, 幹事, Society - Apr. 2019 - Mar. 2023
日本応用数理学会 JSIAM Letters, 編集委員, Society - Apr. 2019 - Mar. 2023
情報処理学会 ACS論文誌, 編集委員, Society - 2022 - 2022
PDSEC ‘22 Program Committee, member, Others - 2022
SC22 Program Committee, member, Society - 2022
HPC Asia 2022 Organizing Committee, Poster chair, Others - 2022
IPDPS2022 Program Committee, member, Others - Apr. 2019 - Mar. 2021
日本応用数理学会 若手の会, 幹事, Society - 2021 - 2021
MCSoC-21：Special Session ATMG Program Committee, Program Vice-Chair, Others - 2021
PDSEC ‘21 Program Committee, member, Others - 2021
iWAPT2021 Program Committee, member, Others - 2021
xSIG2021 プログラム委員会, 委員, Others - 2021
IHPCES2021 Program committee, member, Others - Apr. 2016 - Mar. 2020
情報処理学会 ハイパフォーマンスコンピューティング研究会, 運営委員, Society - 2020
xSIG2020 プログラム委員会, 委員, Others - 2020
IHPCES2020 Program committee, member, Others - 2020
ICPP2020 Program Committee, member, Others - 2020
HPC Asia 2020 Organizing Committee, Publicity chair, Others - 2020
PDSEC'20 Program Committee, member, Others - 2020
iWAPT2020 Program Committee, member, Others - Apr. 2017 - Mar. 2019
日本応用数理学会 若手の会, 主査, Society - 2019
MCSoC-19: Special Session ATMG Program Committee, member, Others - 2019
ICPP2019 Program Committee, member, Others - 2019
PDSEC'19 Program Committee, member, Others - 2019
iWAPT2019 Program Committee, member, Others - 2018
MCSoC-18: Special Session ATMG Program Committee, member, Others - 2018
PDSEC'18 Program Committee, member, Others - 2018
iWAPT2018 Program Committee, member, Others - Apr. 2015 - Mar. 2017
日本応用数理学会 若手の会, 幹事, Society - 2017
MCSoC-17: Special Session ATMG Program Committee, Chair, Others - 2017
PDSEC'17 Program Committee, member, Others - 2017
iWAPT2017 Program Committee, member, Others - 2017
HPCS2017 プログラム委員会, 委員, Society - 2016
iWAPT2016 Program Committee, member, Others - 2016
HPCS2016 プログラム委員会, 副委員長（広報・ポスター担当）, Society - 2015
EPASA2015 Program committee, vice chair, Others - 2015
iWAPT2015 Program Committee, member, Others - 2015
HPCS2015 プログラム委員会, 委員, Society
■Research activity information
Awards
- Aug. 2023, 情報処理学会シンポジウム xSIG2013, Outstanding Research Award
連立一次方程式の求解を前提とした大規模疎行列の条件数推定
工藤 侑也;深谷 猛;岩下 武史, Japan society, Iceland - Dec. 2022, PDCAT2022 Best Paper Award
Distributed Parallel Tall-Skinny QR factorization: Performance Evaluation of Various Algorithms on Various Systems
Takeshi Fukaya, 33945505 - Mar. 2019, IPSJ, IPSJ Yamashita SIG Research Award
タイルレベルの並列処理を可能とする時空間タイリング手法を用いた3次元FDTDカーネルの実装と性能評価
Takeshi FUkaya, Japan society - May 2018, IPSJ Symposium xSIG2018, Best Research Award
Enhancement of Algebraic Block Multi-Color Ordering for ILU Preconditioning and Its Performance Evaluation in Preconditioned GMRES Solver
Senxi Li;Takeshi Iwashita;Takeshi Fukaya, Japan society - Mar. 2010, 名古屋大学, 学術奨励賞
深谷猛, Others - Jun. 2009, EASIAM, 2009 EASIAM Student Paper Competition 2nd Prize
A Dynamic Programming Approach to Optimizing the Blocking Strategy for the Householder QR Decomposition
Fukaya Takeshi, International society - Jan. 2009, 2009年ハイパフォーマンスコンピューティングと計算科学シンポジウム（HPCS2009）, 最優秀論文賞
正方行列向け特異値分解のCUDA による高速化
深谷 猛;山本 有作;畝山 多加志;中村 佳正, Japan society
Papers
- An integer arithmetic-based AMG preconditioned FGMRES solver
Kengo Suzuki, Takeshi Fukaya, Takeshi Iwashita
ACM Transactions on Mathematical Software, Association for Computing Machinery (ACM), 18 Nov. 2024, [Peer-reviewed]
Scientific journal, We consider solving a sparse linear system using integer (fixed-point) arithmetic. Integer arithmetic has attracted attention in scientific computing because of its high computational efficiency. Furthermore, considering the current circumstances of hardware development, integer arithmetic is expected to become increasingly important. Nevertheless, integer arithmetic has not been widely used for solving linear systems because it lacks robustness against overflow and underflow, making it hard to solve practical problems. Thus, we propose a new integer-based implementation framework for the flexible GMRES (FGMRES) method, which enables integer-based solvers to solve linear systems with the same accuracy as conventional floating-point solvers. In addition, we propose an integer-only algebraic multigrid preconditioner. Combining it with the integer-based FGMRES framework, we develop an integer-based solver. Numerical experiments on CPUs showed that the developed integer-based solver has a comparable convergence rate to floating-point solvers. We also found the test cases where the integer-based solver runs faster than the floating-point solvers. - A Cholesky QR type algorithm for computing tall-skinny QR factorization with column pivoting
Takeshi Fukaya, Yuji Nakatsukasa, Yusaku Yamamoto
2024 IEEE International Parallel and Distributed Processing Symposium (IPDPS), 63, 75, IEEE, 27 May 2024, [Peer-reviewed], [Lead author, Corresponding author]
English, International conference proceedings - Subspace Correction Preconditioning for Solving a Sequence of Asymmetric Linear Systems Using the Bi-CGSTAB Method
Hirotoshi Tamori, Takeshi Fukaya, Takeshi Iwashita
Journal of Information Processing, 31, 875, 884, Information Processing Society of Japan, Dec. 2023, [Peer-reviewed]
English, Scientific journal - Numerical Behavior of Mixed Precision Iterative Refinement Using the BiCGSTAB Method
Yingqi Zhao, Takeshi Fukaya, Takeshi Iwashita
Journal of Information Processing, 31, 860, 874, Information Processing Society of Japan, Dec. 2023, [Peer-reviewed]
English, Scientific journal - Convergence acceleration of preconditioned conjugate gradient solver based on error vector sampling for a sequence of linear systems
Takeshi Iwashita, Kota Ikehara, Takeshi Fukaya, Takeshi Mifune
Numerical Linear Algebra with Applications, 30, 6, Wiley, 31 May 2023, [Peer-reviewed]
English, Scientific journal, Abstract
In this article, we focus on solving a sequence of linear systems that have identical (or similar) coefficient matrices. For this type of problem, we investigate subspace correction (SC) and deflation methods, which use an auxiliary matrix (subspace) to accelerate the convergence of the iterative method. In practical simulations, these acceleration methods typically work well when the range of the auxiliary matrix contains eigenspaces corresponding to small eigenvalues of the coefficient matrix. We develop a new algebraic auxiliary matrix construction method based on error vector sampling in which eigenvectors with small eigenvalues are efficiently identified in the solution process. We use the generated auxiliary matrix for convergence acceleration in the following solution step. Numerical tests confirm that both SC and deflation methods with the auxiliary matrix can accelerate the solution process of the iterative solver. Furthermore, we examine the applicability of our technique to the estimation of the condition number of the coefficient matrix. We also present the algorithm of the preconditioned conjugate gradient method with condition number estimation. - Distributed Parallel Tall-Skinny QR Factorization: Performance Evaluation of Various Algorithms on Various Systems
Takeshi Fukaya
Parallel and Distributed Computing, Applications and Technologies, 275, 287, Springer Nature Switzerland, 08 Apr. 2023, [Peer-reviewed], [Lead author, Corresponding author]
English, International conference proceedings, 33945505 - A novel ILU preconditioning method with a block structure suitable for SIMD vectorization
Kengo Suzuki, Takeshi Fukaya, Takeshi Iwashita
Journal of Computational and Applied Mathematics, 419, 114687, 114687, Elsevier BV, Feb. 2023, [Peer-reviewed]
English, Scientific journal - A New AINV Preconditioner for the CG Method in Hybrid CPU-GPU Computing Environment
Kengo Suzuki, Takeshi Fukaya, Takeshi Iwashita
Journal of Information Processing, 30, 755, 765, Information Processing Society of Japan, Oct. 2022, [Peer-reviewed]
English, Scientific journal - Numerical Investigation into the Mixed Precision GMRES(<i>m</i>) Method Using FP64 and FP32
Yingqi Zhao, Takeshi Fukaya, Linjie Zhang, Takeshi Iwashita
Journal of Information Processing, 30, 525, 537, Information Processing Society of Japan, Aug. 2022, [Peer-reviewed]
English, Scientific journal - Performance prediction of massively parallel computation by Bayesian inference
Hisashi Kohashi, Harumichi Iwamoto, Takeshi Fukaya, Yusaku Yamamoto, Takeo Hoshi
JSIAM Letters, 14, 13, 16, The Japan Society for Industrial and Applied Mathematics, 2022, [Peer-reviewed]
English, Scientific journal - An Integer Arithmetic-Based Sparse Linear Solver Using a GMRES Method and Iterative Refinement
Takeshi Iwashita, Kengo Suzuki, Takeshi Fukaya
2020 IEEE/ACM 11th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems (ScalA), 1, 8, IEEE, Nov. 2020, [Peer-reviewed], [International Magazine]
International conference proceedings - Hierarchical block multi-color ordering: a new parallel ordering method for vectorization and parallelization of the sparse triangular solver in the ICCG method
Takeshi Iwashita, Senxi Li, Takeshi Fukaya
CCF Transactions on High Performance Computing, 2, 2, 84, 97, Springer Science and Business Media LLC, Jun. 2020, [Peer-reviewed], [International Magazine]
Scientific journal,Abstract In this paper, we propose a new parallel ordering method to vectorize and parallelize the sparse triangular solver, which is called hierarchical block multi-color ordering. In this method, the parallel forward and backward substitutions can be vectorized while preserving the advantages of block multi-color ordering, that is, fast convergence and fewer thread synchronizations. To evaluate the proposed method in a parallel ICCG (Incomplete Cholesky Conjugate Gradient) solver, numerical tests were conducted using seven test matrices on three types of computational nodes. The numerical results indicate that the proposed method outperforms the conventional block and nodal multi-color ordering methods in 18 out of 21 test cases, which confirms the effectiveness of the method. - Effect of Mixed Precision Computing on H-Matrix Vector Multiplication in BEM Analysis
Rise Ooi, Takeshi Iwashita, Takeshi Fukaya, Akihiro Ida, Rio Yokota
Proceedings of the International Conference on High Performance Computing in Asia-Pacific Region, 92, 101, ACM, 15 Jan. 2020, [Peer-reviewed]
English, International conference proceedings - Shifted Cholesky QR for Computing the QR Factorization of Ill-Conditioned Matrices
Takeshi Fukaya, Ramaseshan Kannan, Yuji Nakatsukasa, Yusaku Yamamoto, Yuka Yanagisawa
SIAM Journal on Scientific Computing, 42, 1, A477, A503, Society for Industrial & Applied Mathematics (SIAM), Jan. 2020, [Peer-reviewed], [Lead author], ［Internationally co-authored］, [International Magazine]
Scientific journal - EigenKernel A middleware for parallel generalized eigenvalue solvers to attain high scalability and usability
Kazuyuki Tanaka, Hiroto Imachi, Tomoya Fukumoto, Akiyoshi Kuwata, Yuki Harada, Takeshi Fukaya, Yusaku Yamamoto, Takeo Hoshi
Japan Journal of Industrial and Applied Mathematics, 36, 2, 719, 742, Springer Science and Business Media LLC, Jul. 2019, [Peer-reviewed], [International Magazine]
Scientific journal - An investigation into the impact of the structured QR kernel on the overall performance of the TSQR algorithm
Takeshi Fukaya
Proceedings of the International Conference on High Performance Computing in Asia-Pacific Region, 81, 90, ACM, 14 Jan. 2019, [Peer-reviewed], [Lead author, Corresponding author], [International Magazine]
International conference proceedings - Enhancement of Algebraic Block Multi-Color Ordering for ILU Preconditioning and Its Performance Evaluation in Preconditioned GMRES Solver
Senxi Li, Takeshi Iwashita, Takeshi Fukaya
Journal of Information Processing, 27, 201, 210, Information Processing Society of Japan, 2019, [Peer-reviewed], [Domestic magazines]
Scientific journal - A Case Study on Modeling the Performance of Dense Matrix Computation: Tridiagonalization in the EigenExa Eigensolver on the K Computer
Takeshi Fukaya, Toshiyuki Imamura, Yusaku Yamamoto
2018 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), 1113, 1122, IEEE, May 2018, [Peer-reviewed], [Lead author, Corresponding author], [International Magazine]
International conference proceedings - Time-space tiling with tile-level parallelism for the 3D FDTD method
Takeshi Fukaya, Takeshi Iwashita
Proceedings of the International Conference on High Performance Computing in Asia-Pacific Region, 116, 126, ACM, 28 Jan. 2018, [Peer-reviewed], [Lead author, Corresponding author], [International Magazine]
International conference proceedings - On Constructing Cost Models for Online Automatic Tuning Using ATMathCoreLib: Case Studies through the SVD Computation on a Multicore Processor
Seiji Nagashima, Takeshi Fukaya, Yusaku Yamamoto
2016 IEEE 10th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSOC), 345, 352, IEEE, Sep. 2016, [Peer-reviewed], [International Magazine]
English, International conference proceedings, We consider the problem of online automatic tuning. In this setting, we execute the target program with some tuning parameters N times, where N is given, while optimizing the parameters to minimize some objective function such as the total execution time. Thus we have to choose the parameters for each execution by taking into account the trade-off between exploration and exploitation. The ATMathCoreLib library developed by Suda is a set of software that solves this problem. To model the performance of the target software, ATMathCoreLib uses a linear statistical model, and its basis functions must be provided by the user.
In this paper, we investigate how to choose the basis functions appropriately, using the singular value decomposition of a square matrix as an example. We consider three cases, namely, (I) when the performance characteristics of the target problem are well understood by the user, (II) when the tuning parameter has a complicated structure, as occurs in the case of simultaneous selection of an algorithm and its parameter, and (III) when the performance characteristics of the target problem are not known to the user. The results of using ATMathCoreLib with different basis functions for each case are given. They help one understand the tuning by ATMathCoreLib and contribute to the progress of ATMathCoreLib. - Application and Evaluation of Various Communication Avoiding Techniques for the Conjugate Gradient Method
熊谷洋佑, 藤井昭宏, 田中輝雄, 深谷猛, 須田礼仁
情報処理学会論文誌トランザクション コンピューティングシステム(Web), 9, 3, 1, 13, 04 Aug. 2016, [Peer-reviewed], [Domestic magazines]
Japanese, Scientific journal, The performance of supercomputers improves as the number of cores increases. The conjugate gradient (CG) method is useful for solving large and sparse linear systems. It has been pointed out that collective communication needed for calculating inner products becomes serious bottleneck when executing the CG method on massively parallel systems. Recently, the Chebyshev basis CG (CBCG) method, a variant of the Communication-avoiding CG method, has been proposed. In this paper, we reduced collective communication of CBCG method (CBCGR) and applied Matrix Powers Kernel (MPK) for CBCGR method. We then measured the execution time of these methods for 2D and 3D Poisson problems using OpenMP/MPI hybrid parallel model on the FX10 (oakleaf-fx) supercomputer system. For the 2D-Poisson problem, the CBCGR and CBCGR-MPK methods are faster than the CG and CBCG methods when the number of processes is sufficiently large. For the 3D-Poisson problem, the CBCGR method is faster than the CG and CBCG methods when the number of processes is sufficeint large. - Performance Analysis of the Chebyshev Basis Conjugate Gradient Method on the K Computer
Yosuke Kumagai, Akihiro Fujii, Teruo Tanaka, Yusuke Hirota, Takeshi Fukaya, Toshiyuki Imamura, Reiji Suda
Parallel Processing and Applied Mathematics, 9573, 74, 85, Springer International Publishing, 2016, [Peer-reviewed], [International Magazine]
In book - CAHTR: Communication-avoiding householder TRidiagonalization
Toshiyuki Imamura, Takeshi Fukaya, Yusuke Hirota, Susumu Yamada, Masahiko Machida
Advances in Parallel Computing, 27, 381, 390, 2016, [Peer-reviewed], [International Magazine]
International conference proceedings, © 2016 The authors and IOS Press. The present paper describes an efficient communication optimization technique for Householder tridiagonalization called CAHTR and evaluates its parallel performance. CAHTR is intended to reduce the number of problems in collective communication, especially MPI Allreduce operations. We demonstrate the optimal version of CAHTR(3) compared with a naive implementation CAHTR(0). The CAHTR algorithms are evaluated on the K supercomputer system, and speedup exceeds x1.4 for the case of N = 5000 and P = 1024. - Roundoff error analysis of the CholeskyQR2 algorithm in an oblique inner product
Yusaku Yamamoto, Yuji Nakatsukasa, Yuka Yanagisawa, Takeshi Fukaya
JSIAM Letters, 8, 5, 8, The Japan Society for Industrial and Applied Mathematics, 2016, [Peer-reviewed], [Domestic magazines]
English, Scientific journal, The Cholesky QR algorithm is an ideal QR decomposition algorithm for high performance computing, but known to be unstable. We present error analysis of the Cholesky QR algorithm in an oblique inner product defined by a positive definite matrix, and show that by repeating the algorithm twice (called CholeskyQR2), its stability is greatly improved. - Performance Evaluation of the Eigen Exa Eigensolver on Oakleaf-FX: Tridiagonalization Versus Pentadiagonalization
Takeshi Fukaya, Toshiyuki Imamura
2015 IEEE International Parallel and Distributed Processing Symposium Workshop, IEEE, May 2015, [Peer-reviewed], [Lead author, Corresponding author], [International Magazine]
International conference proceedings - ROUNDOFF ERROR ANALYSIS OF THE CHOLESKYQR2 ALGORITHM
Yusaku Yamamoto, Yuji Nakatsukasa, Yuka Yanagisawa, Takeshi Fukaya
ELECTRONIC TRANSACTIONS ON NUMERICAL ANALYSIS, 44, 306, 326, KENT STATE UNIVERSITY, 2015, [Peer-reviewed], [International Magazine]
English, Scientific journal, We consider the QR decomposition of an m x n matrix X with full column rank, where m >= n. Among the many algorithms available, the Cholesky QR algorithm is ideal from the viewpoint of high performance computing since it consists entirely of standard level 3 BLAS operations with large matrix sizes, and requires only one reduce and broadcast in parallel environments. Unfortunately, it is well-known that the algorithm is not numerically stable and the deviation from orthogonality of the computed Q factor is of order O((kappa(2)(X))(2) u), where kappa(2)(X) is the 2-norm condition number of X and u is the unit roundoff. In this paper, we show that if the condition number of X is not too large, we can greatly improve the stability by iterating the Cholesky QR algorithm twice. More specifically, if kappa(2)(X) is at most O(u(-1/2)), both the residual and deviation from orthogonality are shown to be of order 0(u). Numerical results support our theoretical analysis. - Performance Analysis of the Householder-Type Parallel Tall-Skinny QR Factorizations Toward Automatic Algorithm Selection
Takeshi Fukaya, Toshiyuki Imamura, Yusaku Yamamoto
Lecture Notes in Computer Science, 8969, 269, 283, Springer International Publishing, 2015, [Peer-reviewed], [Lead author, Corresponding author], [International Magazine]
In book - CholeskyQR2: A Simple and Communication-Avoiding Algorithm for Computing a Tall-Skinny QR Factorization on a Large-Scale Parallel System
Takeshi Fukaya, Yuji Nakatsukasa, Yuka Yanagisawa, Yusaku Yamamoto
2014 5th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems, 31, 38, IEEE, Nov. 2014, [Peer-reviewed], [Lead author, Corresponding author], [International Magazine]
International conference proceedings - Performance Optimization for the Blocked Householder QR Decomposition Using the Dynamic Programming
深谷 猛, 山本 有作, 張 紹良
情報処理学会論文誌 論文誌トランザクション, 2011, 2, 146, 157, 情報処理学会, Apr. 2012, [Peer-reviewed], [Lead author, Corresponding author], [Domestic magazines]
Japanese, Scientific journal - Acceleration of Hessenberg Reduction for Nonsymmetric Eigenvalue Problems in a Hybrid CPU-GPU Computing Environment.
Jun-ichi Muramatsu, Takeshi Fukaya, Shao-Liang Zhang, Kinji Kimura, Yusaku Yamamoto
IJNC, 1, 2, 132, 143, 2011, [Peer-reviewed], [International Magazine]
Scientific journal - Differential qd algorithm for totally nonnegative Hessenberg matrices: introduction of origin shifts and relationship with the discrete hungry Lotka-Volterra system
Yusaku Yamamoto, Takeshi Fukaya
JSIAM Letters, 2, 69, 72, The Japan Society for Industrial and Applied Mathematics, 2010, [Peer-reviewed], [Domestic magazines]
English, Scientific journal, We propose an approach for introducing the origin shift into the multiple dqd algorithm for computing the eigenvalues of a totally nonnegative matrix. Numerical experiments show that the shift speeds up the convergence while retaining the accuracy of the computed eigenvalue. - Acceleration of the Singular Value Decomposition Algorithm for Square Matrices Using CUDA
深谷 猛, 山本 有作, 畝山 多加志, 中村 佳正
情報処理学会論文誌コンピューティングシステム（ACS）, 2, 2, 98, 109, 情報処理学会, 02 Jul. 2009, [Peer-reviewed], [Lead author, Corresponding author], [Domestic magazines]
Japanese, Scientific journal, In this paper, we report the result of acceleration of computing the singular value decomposition (SVD) for a square matrix using CUDA, which is an integrated development environment for GPGPU. Computing of the SVD for a square matrix consists of the following three parts: bidiagonalization of the input matrix, the SVD of the bidiagonal matrix, and inverse transformation. Among them, we accelerate the first and the third step using GPU. This is because it is easy to use the CUBLAS, the BLAS library provided in CUDA, in these two steps. Through simple performance prediction, we assessed that the Bischof's method, in which bidiagonalization can be computed with matrix multiplications, is effective for computation using GPU. Therefore we implemented the algorithm for the SVD based on such method. When computing the SVD of a 5,120×5,120 matrix, we obtained about four times speedup using a GPU over using only a CPU (Intel Core2 Duo, 1.86 GHz, using 2 cores). - An efficient bidiagonalization algorithm for combined CPU-accelerator environments
Yusaku Yamamoto, Takeshi Fukaya, Takashi Uneyama, Yoshimasa Nakamura
Proceedings of the IASTED International Conference on Parallel and Distributed Computing and Networks, PDCN 2009, 121, 126, 2009
English, International conference proceedings, In computing the singular values of a square matrix, transformation of the input matrix to bidiagonal form accounts for most of the computation time. In this paper, we consider speeding up this process using a combination of CPU and floating-point accelerator. As an algorithm for bidiagonalization, we can use the conventional Householder's method or Bischof's two-phase algorithm, which can use the level-3 BLAS efficiently. We can also choose to store the whole matrix in the CPU memory or in the on-board memory of the accelerator. So there are four possible strategies. We investigate the advantages and disadvantages of each strategy and construct an analytical performance model for each of them. Using the models, we predict the performance of bidiagonalzation on the CSX600 accelerator and show that it is the best to achieve high performance to use Bischof's algorithm with the matrix stored in the on-board memory. This conclusion should hold for many other accelerators with similar performance characteristics. - Differential qd algorithm for totally nonnegative band matrices: convergence properties and error analysis
Yusaku Yamamoto, Takeshi Fukaya
JSIAM Letters, 1, 56, 59, The Japan Society for Industrial and Applied Mathematics, 2009, [Peer-reviewed], [Domestic magazines]
English, Scientific journal, We analyze convergence properties and numerical properties of the differential qd algorithm generalized for totally nonnegative band matrices. In particular, we show that the algorithm is globally convergent and can compute all eigenvalues to high relative accuracy. - A dynamic programming approach to optimizing the blocking strategy for the Householder QR decomposition
Takeshi Fukaya, Yusaku Yamamoto, Shao-Liang Zhang
2008 IEEE International Conference on Cluster Computing, 402, 410, IEEE, Sep. 2008, [Peer-reviewed], [Lead author, Corresponding author], [International Magazine]
International conference proceedings - Acceleration of the Singular Value Decomposition Algorithm for Rectangular Matrices with a Floating-point Coprocessor
深谷猛, 山本有作, 畝山多加志, 堀玄, 梅野健
情報処理学会論文誌, 48, SIG8(ACS18), 31, 43, May 2007, [Peer-reviewed], [Lead author, Corresponding author], [Domestic magazines]
Japanese, Scientific journal - Accelerating the Singular Value Decomposition of Rectangular Matrices with the CSX600 and the Integrable SVD
Yusaku Yamamoto, Takeshi Fukaya, Takashi Uneyama, Masami Takata, Kinji Kimura, Masashi Iwasaki, Yoshimasa Nakamura
Lecture Notes in Computer Science, 4671, 340, 345, Springer Berlin Heidelberg, 2007, [Peer-reviewed], [International Magazine]
In book
Other Activities and Achievements
- Accelerating the SpMV kernel on standard CPUs by exploiting the partially diagonal structures
Takeshi Fukaya, Koki Ishida, Akie Miura, Takeshi Iwashita, Hiroshi Nakashima, CoRR, abs/2105.04937, 11 May 2021, [Lead author, Corresponding author]
Sparse Matrix Vector multiplication (SpMV) is one of basic building blocks in
scientific computing, and acceleration of SpMV has been continuously required.
In this research, we aim for accelerating SpMV on recent CPUs for sparse
matrices that have a specific sparsity structure, namely a diagonally
structured sparsity pattern. We focus a hybrid storage format that combines the
DIA and CSR formats, so-called the HDC format. First, we recall the importance
of introducing cache blocking techniques into HDC-based SpMV kernels. Next,
based on the observation of the cache blocked kernel, we present a modified
version of the HDC formats, which we call the M-HDC format, in which partial
diagonal structures are expected to be more efficiently picked up. For these
SpMV kernels, we theoretically analyze the expected performance improvement
based on performance models. Then, we conduct comprehensive experiments on
state-of-the-art multi-core CPUs. By the experiments using typical matrices, we
clarify the detailed performance characteristics of each SpMV kernel. We also
evaluate the performance for matrices appearing in practical applications and
demonstrate that our approach can accelerate SpMV for some of them. Through the
present paper, we demonstrate the effectiveness of exploiting partial diagonal
structures by the M-HDC format as a promising approach to accelerating SpMV on
CPUs for a certain kind of practical sparse matrices. - 縦長行列のQR分解に対する各種アルゴリズムの比較：Oakforest-PACS上での性能評価
深谷猛, 東京大学情報基盤センター スーパーコンピューティングニュース, 22, 6, 28, 39, Dec. 2020, [Lead author, Corresponding author], [Domestic magazines]
Introduction research institution - ブロックに基づくfill-in選択手法を利用したILU-GMRESソルバ
鈴木 謙吾, 深谷 猛, 岩下 武史, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2020-HPC-177, 20, 1, 7, Dec. 2020, [Domestic magazines]
Summary national conference - White Paper from Workshop on Large-scale Parallel Numerical Computing Technology (LSPANC 2020): HPC and Computer Arithmetic toward Minimal-Precision Computing
Roman Iakymchuk, Daichi Mukunoki, Artur Podobas, Fabienne Jézéquel, Toshiyuki Imamura, Norihisa Fujita, Jens Huthmann, Shuhei Kudo, Yiyu Tan, Jens Domke, Kai Torben Ohlhus, Takeshi Fukaya, Takeo Hoshi, Yuki Murakami, Maho Nakata, Takeshi Ogita, Kentaro Sano, Taisuke Boku, https://arxiv.org/abs/2004.04628, abs/2004.04628, 09 Apr. 2020, ［Internationally co-authored］
In numerical computations, precision of floating-point computations is a key
factor to determine the performance (speed and energy-efficiency) as well as
the reliability (accuracy and reproducibility). However, precision generally
plays a contrary role for both. Therefore, the ultimate concept for maximizing
both at the same time is the minimal-precision computing through
precision-tuning, which adjusts the optimal precision for each operation and
data. Several studies have been already conducted for it so far (e.g.
Precimoniuos and Verrou), but the scope of those studies is limited to the
precision-tuning alone. Hence, we aim to propose a broader concept of the
minimal-precision computing system with precision-tuning, involving both
hardware and software stack.
In 2019, we have started the Minimal-Precision Computing project to propose a
more broad concept of the minimal-precision computing system with
precision-tuning, involving both hardware and software stack. Specifically, our
system combines (1) a precision-tuning method based on Discrete Stochastic
Arithmetic (DSA), (2) arbitrary-precision arithmetic libraries, (3) fast and
accurate numerical libraries, and (4) Field-Programmable Gate Array (FPGA) with
High-Level Synthesis (HLS).
In this white paper, we aim to provide an overview of various technologies
related to minimal- and mixed-precision, to outline the future direction of the
project, as well as to discuss current challenges together with our project
members and guest speakers at the LSPANC 2020 workshop;
https://www.r-ccs.riken.jp/labs/lpnctrt/lspanc2020jan/. - ランタイムシステムを用いたマルチフロンタルコレスキー分解の開発
中野 智輝, 横川 三津夫, 深谷 猛, 山本 有作, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2020-HPC-173, 10, 1, 14, Mar. 2020, [Domestic magazines]
Summary national conference - テンソル分解におけるMTTKRPのスレッド並列化に関する考察
深谷猛, 計算工学講演会論文集, 24, May 2019, [Lead author, Corresponding author], [Domestic magazines]
Summary national conference - 緩和型スーパーノードマルチフロンタル法の最適な緩和パラメータについて
中野 智輝, 横川 三津夫, 深谷 猛, 山本 有作, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2018-HPC-167, 25, 1, 8, Dec. 2018, [Domestic magazines]
数値シミュレーションにおける多くの問題は，偏微分方程式を離散化して得られる連立一次方程式を解く問題に帰着される．そして，多くの場合，連立一次方程式を解く時間は全体のシミュレーション時間の大部分を占める．よって，連立一次方程式を高速に解くことは非常に重要である．本研究では，正定値対称行列に適用できるコレスキー分解を扱う．疎行列に対して，コレスキー分解を行う手法はいくつかあるが，本稿では，緩和型スーパーノードマルチフロンタル法を用いた．同手法では，2 つのスーパーノードを融合する際に非零と見なす零要素数の上限である緩和パラメータが性能に大きな影響を与える。そこで，このパラメータの最適値を求めることを目的として，Intel Xeon (Ivy Bridge-EX) とIntel Xeon Phi(Knights Landing, KNL) のそれぞれ1 コ, Information Processing Society of Japan, Japanese, Summary national conference - Chebyshev基底通信削減CG法のマルチコア・メニーコア計算環境における性能評価
大島聡史, 藤井昭宏, 田中輝雄, 深谷猛, 須田礼仁, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2018-HPC-165, 17, 1, 9, Jul. 2018, [Domestic magazines]
Summary national conference - Knights LandingにおけるTilied3D FDTDカーネルの性能評価
深谷猛, 岩下武史, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2018-HPC-164, 6, 1, 9, May 2018, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - One‐way dissectionオーダリングによる連立一次方程式の直接解法の並列化
中野智輝, 横川三津夫, 深谷猛, 山本有作, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2017-HPC-162, 19, 1, 10, 11 Dec. 2017, [Domestic magazines]
Japanese, Summary national conference - タイルレベルの並列処理を可能とする時空間タイリング手法を用いた3次元FDTDカーネルの実装と性能評価
深谷猛, 岩下武史, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2017-HPC-160, 35, 1, 11, 19 Jul. 2017, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - Acceleration of sparse matrix vector multiplication by exploiting stencil structure
深谷猛, 三浦瑛絵, 岩下武史, 計算工学講演会論文集, 22, 4p, 31 May 2017, [Lead author, Corresponding author], [Domestic magazines]
日本計算工学会, Japanese, Summary national conference - 大規模並列計算機における連立一次方程式の精度保証付き数値計算に対する性能評価
森倉悠介, 椋木大地, 深谷猛, 山中脩也, 大石進一, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2016-HPC-157, 1, 1, 7, 14 Dec. 2016, [Domestic magazines]
Japanese, Summary national conference - 1次元分散型のCAQRアルゴリズムの性能評価とパネルサイズの自動チューニングに向けた検討
深谷猛, 深谷猛, 深谷猛, 山本有作, 山本有作, 今村俊幸, 今村俊幸, 計算工学講演会論文集, 20, 08 Jun. 2015, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - QR分解に対する通信回避型アルゴリズムと自動チューニング
深谷猛, 計算工学, 20, 2, 3247, 3250, 30 Apr. 2015, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Introduction scientific journal - FX10 4800ノードを用いた通信削減型QR分解アルゴリズムの性能評価
深谷猛, 東京大学情報基盤センター スーパーコンピューティングニュース, 16, 4, 11, 20, Jul. 2014, [Lead author, Corresponding author], [Domestic magazines]
Introduction research institution - 密行列固有値計算における通信回避(CA)と通信隠蔽(CH)について
今村俊幸, 廣田悠輔, 深谷猛, 山田進, 町田昌彦, 計算工学講演会論文集, 19, 11 Jun. 2014, [Domestic magazines]
Japanese, Summary national conference - TSQRで生じる特殊な構造を持ったQR分解に対する自動チューニングの検討
深谷猛, 今村俊幸, 計算工学講演会論文集, 19, 11 Jun. 2014, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - 通信削減アルゴリズムCAQRのRSDFTの直交化処理への適用と評価
片桐孝洋, 高山恒一, 米村崇, 熊洞宏樹, 猪貝光祥, 北上純一, 江口義之, 深谷猛, 山本有作, 岩田潤一, 内田和之, 大島聡史, 中島研吾, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2014-HPC-144, 3, 1, 6, 19 May 2014, [Domestic magazines]
Japanese, Summary national conference - FX10 4800ノードを用いた密行列向け固有値ソルバEigenExaの性能評価
深谷 猛, 今村 俊幸, 東京大学情報基盤センター スーパーコンピューティングニュース, 16, 3, 20, 27, May 2014, [Lead author, Corresponding author], [Domestic magazines]
Introduction research institution - 超並列環境向け固有値計算プログラムの性能予測モデルの開発（続）
深谷猛, 東京大学情報基盤センター スーパーコンピューティングニュース, 16, 1, 21, 28, Jan. 2014, [Lead author, Corresponding author], [Domestic magazines]
Introduction research institution - 超並列環境向け固有値計算プログラムの性能予測モデルの開発
深谷猛, 東京大学情報基盤センター スーパーコンピューティングニュース, 15, 6, 33, 43, Nov. 2013, [Lead author, Corresponding author], [Domestic magazines]
Introduction research institution - 超並列環境における密行列計算プログラムの性能モデリングに向けた検討
深谷猛, 今村俊幸, 山本有作, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2013-HPC-140, 41, 1, 8, 24 Jul. 2013, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - 超並列環境における縦長行列のQR分解に対する自動チューニングの検討
深谷猛, 山本有作, 計算工学講演会論文集, 18, 19 Jun. 2013, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - 京における密行列固有値ソルバEigen-Kの性能評価と性能モデリング
深谷猛, 今村俊幸, 山本有作, 先進的計算基盤システムシンポジウム論文集, 2013, 132, 133, 15 May 2013
Japanese - Performance Optimization for the Blocked Householder QR Decomposition Using the Dynamic Programming
深谷 猛, 山本 有作, 張 紹良, 情報処理学会論文誌コンピューティングシステム（ACS）, 4, 4, 146, 157, 05 Oct. 2011
密行列計算においては，高性能化のためにアルゴリズムのブロック化が必須である．その際に，ブロック化の方法次第で性能が大きく変化するため，その最適化が重要な課題となっている．しかしながら，ブロック化の自由度が大きいため，従来は限定された範囲内で最適化を行うことがほとんどである．本論文では，QR 分解アルゴリズムを対象として，二分木を使うことで従来より格段に広いクラスのブロック化の方法を系統的に扱い，その中から動的計画法により最適なブロック化の方法を決定する手法を提案する．数値実験の結果，提案手法がブロック分割法に対する自動チューニング手法として有望であることが示された．Blocking techniques are widely used in high performance matrix computations. When using them, it is important to optimize a blocking way, which influences the performance of computations. However, because of the high degree of freedom in blocking techniques, such optimization is generally done in a limited class of blocking ways. In this paper, we propose a framework to determine the efficient blocking way for the algorithm of QR decomposition. In our framework, various kinds of blocking ways are represented systematically with binary trees and an optimal one is determined by dynamic programming. Results of numerical experiments show that our framework has good possibilities in the view of the automatic performance tuning., 情報処理学会, Japanese - QR分解アルゴリズムに対する自動チューニング―性能モデルに関する考察―
深谷猛, 山本有作, ZHANG Shao‐Liang, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2011-HPC-130, 42, 1, 6, 15 Aug. 2011, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - 動的計画法に基づく密行列計算アルゴリズムの再帰的ブロック化
深谷猛, 山本有作, 張紹良, ハイパフォーマンスコンピューティングと計算科学シンポジウム論文集, 2011, 65, 65, 11 Jan. 2011
Japanese - 密行列計算アルゴリズムに対するブロック分割法の最適化と性能評価
深谷猛, 山本有作, ZHANG Shao‐Liang, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2010-HPC-126, 33, 1, 6, 15 Oct. 2010, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - 階層的な性能モデルに基づく行列計算の自動チューニング
山本有作, 深谷猛, 応用数理, 20, 3, 201, 211, 24 Sep. 2010, [Domestic magazines]
Japanese, Introduction scientific journal - Perforamnce Optimization and Evaluation of the Blocking Strategy for the Dense Matrix COmputations
FUKAYA TAKESHI, YAMAMOTO YUSAKU, ZHANG SHAO-LIANG, 研究報告ハイパフォーマンスコンピューティング（HPC）, 2010, 33, 1, 6, 27 Jul. 2010
高性能な行列計算を行う場合，プログラムの性能チューニングが必要不可欠である．我々は基本的な密行列計算が BLAS ルーチンを使って実行される点に着目し，チューニング済みの BLAS ルーチンを効率的に使えるようにプログラムをチューニングすることを目指す．ブロック化されたアルゴリズムにおいて，効率的に BLAS を使うためには行列のブロック分割法を最適化することが重要となる．本稿では，LU 分解のアルゴリズムをブロック化して，ブロック分割法が性能に与える影響を評価し，さらに適切な分割法を決定するための手法の検討を行う．For high performance matrix computations, it is necessity to tune the software. Since basic dense matrix computations consist almost entirely of the BLAS routines, it is important how to tune programs for exploiting the peak performance of optimized BLAS routines. In blocked algorithm, this means how to optimize the partitioning of the target matrix. In this paper, we evaluate and discuss the blocking strategy for the blocked LU decomposition., 情報処理学会, Japanese - ブロックハウスホルダーQR分解の並列計算における自動チューニング手法の検討
深谷猛, 山本有作, ZHANG Shao‐Liang, 情報処理学会研究報告：ハイパフォーマンスコンピューティング, 2009-HPC-121, 18, 1, 7, 15 Oct. 2009, [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - A Study on Automatic Tuning for Parallel Computation of the Blocked Housseholder QR Decomposition
深谷 猛, 山本 有作, 張 紹良, 研究報告ハイパフォーマンスコンピューティング（HPC）, 2009, 18, 1, 7, 28 Jul. 2009
行列計算を並列化する場合，行列ベクトル積や行列乗算などの BLAS ルーチンを並列化する方法と，それらのルーチンをコールする階層で並列化する方法が考えられる．また，行列をブロックに分割して計算を行うことが一般的となっている．そのため，ユーザーは並列化方法とブロック分割法の両者のチューニングを行う必要があるが，自由度が非常に大きいため，効果的なチューニングをすることが難しい．そこで，本稿ではハウスホルダー QR 分解を対象として，自動チューニング手法の検討を行う．In matrix computation, we can parallelize an algorithm by two ways: parallelization of BLAS routines such as matrix-vector multiplication, and parallelization in algorithm levels where BLAS routines are called. In addition, blocking techniques are widely used for matrix computations. Therefore we have many choices when tuning our programs for parallel computers. But it is very difficult for general users to tune their programs effectively. In this paper, we discuss an approach to automatic tuning the algorithm of the blocked Householder QR decomposition., 情報処理学会, Japanese - 正方行列向け特異値分解のCUDAによる高速化
深谷猛, 山本有作, 畝山多加志, 中村佳正, 2009年ハイパフォーマンスコンピューティングと計算科学シンポジウム（HPCS2009）論文集, 107, 114, 15 Jan. 2009, [Peer-reviewed], [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference - 長方行列向け特異値分解の浮動小数点コプロセッサによる高速化
深谷猛, 山本有作, 畝山多加志, 堀玄, 梅野健, 2007年ハイパフォーマンスコンピューティングと計算科学シンポジウム（HPCS2007）論文集, 111, 118, 17 Jan. 2007, [Peer-reviewed], [Lead author, Corresponding author], [Domestic magazines]
Japanese, Summary national conference
Books and other publications
- Sustained Simulation Performance 2018 and 2019
Tomoki Nakano, Mitsuo Yokokawa, Yusaku Yamamoto, Takeshi Fukaya, Affecting the Relaxation Parameter in the Multifrontal Method
Springer, 2020, 9783030391805, 215-224, [Contributor] - 数値線形代数の数理とHPC
櫻井 鉄也, 松尾 宇泰, 片桐 孝洋, 日本応用数理学会, 第6章 固有値・特異値問題における並列計算 6.1 直接法
共立出版, 2018, 9784320019553, 229-249, Japanese, [Contributor] - Software automatic tuning : from concepts to state-of-the-art results
直野, 健, 寺西, 慶太, Cavazos, John, 須田, 礼仁, Dynamic Programming Approaches to Optimizing the Blocking Strategy for Basic Matrix Decompositions
Springer, 2010, 9781441969347, xiv, 377 p., 69-85, English, [Peer-reviewed], [Contributor]
Lectures, oral presentations, etc.
- 低精度演算を用いた線形計算アルゴリズムの研究
深谷 猛
第7回北大・部局横断シンポジウム, 01 Oct. 2021, Oral presentation
オンライン, [Domestic Conference] - ベイズ推定による超並列計算の性能予測
星 健夫, 小橋 恒士, 山本 有作, 深谷 猛
日本応用数理学会2021年度年会, 09 Sep. 2021, 日本応用数理学会, Oral presentation
オンライン, [Domestic Conference] - GPUに適した近似逆行列前処理の簡略化手法
鈴木 謙吾, 深谷 猛, 岩下 武史
日本応用数理学会2021年度年会, 07 Sep. 2021, 日本応用数理学会, Oral presentation
オンライン, Japan, [Domestic Conference] - GMRES(m)法に対する低精度演算・データの積極的導入の可能性に関する検証
深谷 猛, 岩下 武史
日本応用数理学会2021年度年会, 07 Sep. 2021, 日本応用数理学会, Oral presentation
オンライン, Japan, [Domestic Conference] - 最近のマルチコアCPU環境における疎行列ベクトル積の性能に関する一考察
深谷 猛, 岩下 武史, 中島 浩
日本応用数理学会「行列・固有値問題の解法とその応用」研究部会 第31回研究会（SwoPP2021）, 20 Jul. 2021, 日本応用数理学会「行列・固有値問題の解法とその応用」研究部会, Oral presentation
オンライン, Japan, [Domestic Conference] - SIMD演算に適したブロック構造を有する新しいILU分解前処理手法
鈴木 謙吾, 深谷 猛, 岩下 武史
The 5th cross-disciplinary Workshop on Computing Systems, Infrastructures, and Programming (xSIG2021), 19 Jul. 2021, 情報処理学会 ARC/HPC/OS/PRO 各研究会, Oral presentation
オンライン, [Domestic Conference] - Exploiting Lower Precision Computing in the GMRES(m) Method
Takeshi Fukaya, Yingqi Zhao, Takeshi Iwashita
SIAM Conference on Applied Linear Algebra (LA21), 20 May 2021, SIAM, Oral presentation
online, [International presentation] - Exploiting Lower Precision Computing in the GMRES(m) Method
Takeshi Fukaya
2021 Conference on Advanced Topics and Auto Tuning in High-Performance Scientific Computing (ATAT2021), 19 Mar. 2021, Oral presentation
Taoyuan City & online, Taiwan, Province of China, [International presentation] - GMRES(m)法における行列データの低精度化に関する検討
深谷 猛, 岩下 武史
日本応用数理学会 第17回研究部会連合発表会, 04 Mar. 2021, 日本応用数理学会, Oral presentation
オンライン, Japan, [Domestic Conference] - Hierarchical Block Multi-Color Ordering for Vectorization and Parallelization of the ICCG Method
Takeshi Iwashita, Senxi Li, Takeshi Fukaya
SIAM Conference on Computational Science and Engineering (CSE21), 04 Mar. 2021, SIAM, Oral presentation
online, [International presentation] - 低精度・低信頼性演算を活用した数値計算アルゴリズムの創出
深谷 猛
第12回 自動チューニング技術の現状と応用に関するシンポジウム(ATTA2020), 25 Dec. 2020, 自動チューニング研究会, Oral presentation
オンライン, [Domestic Conference] - 縦長行列の列ピボット付きQR分解に対するコレスキーQR型アルゴリズムの検討
深谷 猛, 中務 佑治, 山本 有作
日本応用数理学会2020年度年会, 09 Sep. 2020, 日本応用数理学会, Oral presentation
オンライン, Japan, [Domestic Conference], ［Internationally co-authored］ - Automated Subspace Correction法を前処理とするCGソルバの開発と評価
池原 紘太, 深谷 猛, 岩下 武史
The 4th cross-disciplinary Workshop on Computing Systems, Infrastructures, and Programming (xSIG2020), 29 Jul. 2020, 情報処理学会 ARC/HPC/OS/PRO 各研究会, Oral presentation
オンライン, Japan, [Domestic Conference] - Shifted CholeskyQR3 for High Performance Tall-Skinny QR Factorization
Takeshi Fukaya, Ramaseshan Kannan, Yuji Nakatsukasa, Yusaku Yamamoto, Yuka Yanagisawa
SIAM Conference on Parallel Processing for Scientific Computing (SIAM PP20), 13 Feb. 2020, SIAM, Oral presentation
Seattle, United States, [International presentation], ［Internationally co-authored］ - Investigation into the convergence behavior of the mixed-precision GMRES(m) method using FP64 and FP32
Takeshi Fukaya
Workshop on Large-scale Parallel Numerical Computing Technology (LSPANC 2020 January), 29 Jan. 2020, RIKEN R-CCS, Oral presentation
Kobe, Japan, [International presentation] - Benchmarking Basic Dense Linear Algebra Kernels on the supercomputer Grand Chariot
Takeshi Fukaya
Sapporo Winter HPC Seminar 2020, 24 Jan. 2020, Information Initiative Center, Hokkaido University, Oral presentation
Sapporo, Japan, [International presentation] - HPC視点に基づくテンソル分解アルゴリズムの高性能化
深谷 猛
第11回 自動チューニング技術の現状と応用に関するシンポジウム(ATTA2019), 23 Dec. 2019, 自動チューニング研究会, Oral presentation
東京都, Japan, [Domestic Conference] - 北海道大学情報基盤センター 新スーパーコンピュータシステム利用者からの問い合わせ分析
吉川 潤, 更科 高広, 吉川 浩, 金子 修己, 岩﨑 誠, 折野 神惠, 岩舩 歩美, 深谷 猛, 岩下 武史
大学ICT推進協議会2019年度年次大会（AXIES2019）, 13 Dec. 2019, 一般社団法人 大学ICT推進協議会, Oral presentation
福岡市, Japan, [Domestic Conference] - 3 次元FDTD 法に対する並列処理に適した時空間タイリング手法
深谷 猛
北海道大学共同利用・共同研究拠点アライアンス 部局横断シンポジウム「計算科学が拓く汎分野研究」, 31 Oct. 2019, 北海道大学共同利用・共同研究拠点アライアンス, Oral presentation
札幌市, Japan, [Domestic Conference] - 倍精度と単精度を用いた混合精度GMRES(m)法の収束性に関する実験的評価
深谷 猛, グドール 聖哉, 張 臨傑, 岩下 武史
日本応用数理学会2019年度年会, 03 Sep. 2019, 日本応用数理学会, Oral presentation
東京都, Japan, [Domestic Conference], ［Internationally co-authored］ - Recent progress of the Cholesky QR factorization
Takeshi Fukaya
2019 Mini-Workshop on Computational Science (MWCS2019), 18 Aug. 2019, Oral presentation
Dalian, China, [International presentation] - Mixed-Precision GMRES(m) Method using Double and Single Precision: Experimental Evaluation of its Convergence Properties
Takeshi Fukaya
Sapporo Summer HPC Seminar 2019, 15 Aug. 2019, Information Initiative Center, Hokkaido University, Oral presentation
Sapporo, Japan, [International presentation] - 倍精度と単精度を用いた混合精度 GMRES(m) 法の性能評価
深谷 猛, グドール 聖哉, 張 臨傑, 岩下 武史
第48回数値解析シンポジウム（NAS2019）, 12 Jun. 2019, Oral presentation
福井市, Japan, [Domestic Conference], ［Internationally co-authored］ - ALS法を用いた密テンソルのCP分解におけるMTTKRPの性能評価
深谷 猛
The 3rd cross-disciplinary Workshop on Computing Systems, Infrastructures, and Programming (xSIG2019), 28 May 2019, 情報処理学会 ARC/HPC/OS/PRO 各研究会, Oral presentation
横浜市, Japan, [Domestic Conference] - ベクトル直交化手法に関する最近の進展
深谷 猛
大規模並列数値計算技術に関する研究集会 (LSPANC2019 March), 26 Mar. 2019, 理研 R-CCS, Oral presentation
神戸市, Japan, [Domestic Conference] - Accelerating Multithreaded Linear Solver with Mixed Precision Hierarchical Matrix Computation and Data Structure
Rise Ooi Kok Thong, Takeshi Fukaya, Takeshi Iwashita
日本応用数理学会若手の会主催 第4回学生研究発表会, 03 Mar. 2019, 日本応用数理学会 若手の会, Poster presentation
つくば市, Japan, [Domestic Conference] - High performance QR factorization of ill-conditioned matrices based on the Cholesky QR algorithm
Takeshi Fukaya, Ramaseshan Kannan, Yuji Nakatsukasa, Yusaku Yamamoto, Yuka Yanagisawa
SIAM Conference on Computational Science and Engineering (CSE19), 27 Feb. 2019, SIAM, Oral presentation
Spokane, United States, [International presentation], ［Internationally co-authored］ - Shifted Cholesky QR algorithm for computing the QR factorization of ill-conditioned matrices
Takeshi Fukaya, Ramaseshan Kannan, Yuji Nakatsukasa, Yusaku Yamamoto, Yuka Yanagisawa
2019 Conference on Advanced Topics and Auto Tuning in High-Performance Scientific Computing (ATAT2019), 15 Feb. 2019, Oral presentation
Kaohsiung, Taiwan, Province of China, [International presentation], ［Internationally co-authored］ - 超並列計算に対するベイズ推定型性能予測
原田 祐希, 田中 和幸, 深谷 猛, 山本 有作, 星 健夫
ポスト「京」重点課題（７）「次世代の産業を支える新機能デバイス・ 高性能材料の創成(CDMSI）」第4回シンポジウム, 17 Dec. 2018, Poster presentation
東京都, Japan, [Domestic Conference] - 北海道大学情報基盤センター新スーパーコンピュータシステムの概要
深谷 猛, 岩下 武史, 金子 修己, 折野 神惠, 更科 高広
大学ICT推進協議会2018年度年次大会（AXIES2018）, 21 Nov. 2018, 一般社団法人 大学ICT推進協議会, Oral presentation
札幌市, Japan, [Domestic Conference] - Performance Evaluation of the Shifted Cholesky QR Algorithm for Ill-Conditioned Matrices
Takeshi Fukaya, Ramaseshan Kannan, Yuji Nakatsukasa, Yusaku Yamamoto, Yuka Yanagisawa
SC’18: The International Conference for High Performance Computing, Networking, Storage, and Analysis, 2018, IEEE/ACM, Poster presentation
11 Nov. 2018 - 16 Nov. 2018, Dallas, United States, [International presentation], ［Internationally co-authored］ - コレスキー分解を用いたQR分解の高性能計算手法
深谷 猛
名古屋大学 張研究室 コロキウム, 19 Oct. 2018, 名古屋大学 張研究室, Oral presentation
名古屋市, Japan, [Domestic Conference] - Bayesian Inference Based Performance Prediction For Massively Parallel Numerical Solver
Yuki Harada, Kazuyuki Tanaka, Takeshi Fukaya, Yusaku Yamamoto, Takeo Hoshi
3rd International Symposium on Research and Education of Computational Science (RECS2018), 21 Sep. 2018, The Computational Science Alliance, The University of Tokyo, Poster presentation
Tokyo, Japan, [International presentation] - An overview of various algorithms for computing tall-skinny QR factorization
Takeshi Fukaya, Yusaku Yamamoto
The 37th JSST Annual International Conference on Simulation Technology (JSST2018), 18 Sep. 2018, JAPAN SOCIETY FOR SMILATION TECHNOLOGY, Oral presentation
Muroran, Japan, [International presentation] - High performance multi-threaded ILU-GMRES solver with algebraic block multi-color ordering
Takeshi Iwashita, Senxi Li, Takeshi Fukaya
CoSaS 2018: International Symposium on Computational Science at Scale, 2018, Poster presentation
05 Sep. 2018 - 07 Sep. 2018, Erlangen, Germany, [International presentation] - マルチコア・メニーコア計算機環境におけるChebyshev基底通信削減CG法の性能評価
大島 聡史, 藤井 昭宏, 田中 輝雄, 深谷 猛, 須田 礼仁
日本応用数理学会2018年度年会, 05 Sep. 2018, 日本応用数理学会, Oral presentation
名古屋市, Japan, [Domestic Conference] - 密テンソルに対するALS法の実装方法に関する考察
深谷 猛
日本応用数理学会2018年度年会, 05 Sep. 2018, 日本応用数理学会, Oral presentation
名古屋市, Japan, [Domestic Conference] - シフト付きCholeskyQR法を用いた一般内積空間におけるQR分解の計算
深谷 猛, 中務 佑治, Kannan Ramaseshan, 山本 有作, 柳澤 優香
日本応用数理学会2018年度年会, 04 Sep. 2018, 日本応用数理学会, Poster presentation
名古屋市, Japan, [Domestic Conference] - ベイズ推定を用いた並列数値計算ライブラリの性能予測
原田 祐希, 田中 和幸, 福本 智哉, 深谷 猛, 山本 有作, 星 健夫
日本応用数理学会2018年度年会, 04 Sep. 2018, 日本応用数理学会, Poster presentation
名古屋市, Japan, [Domestic Conference] - QR factorization via Cholesky factorization
Takeshi Fukaya
Sapporo Summer HPC Seminar 2018, 08 Aug. 2018, Information Initiative Center, Hokkaido University, Oral presentation
Sapporo, Japan, [International presentation] - H行列ベクトル積のスレッド並列化における負荷均衡に関する検討
岩下 武史, 川村 卓人, 深谷 猛, 伊田 明弘
日本応用数理学会「行列・固有値問題の解法とその応用」研究部会 第25回研究会（SwoPP2018）, 31 Jul. 2018, 日本応用数理学会「行列・固有値問題の解法とその応用」研究部会, Oral presentation
熊本市, Japan, [Domestic Conference] - ベイズ推定を用いた並列固有値ソルバーの性能予測
田中 和幸, 深谷 猛, 山本 有作, 星 健夫
H30年度 ポスト「京」重点課題（７） 第3回CDMSI研究会, 19 Jul. 2018, Oral presentation
東京都, Japan, [Domestic Conference] - DIA 形式と CRS 形式を組み合わせた Hybrid 形式を用いた疎行列ベクトル積のキャッシュブロッキング
石田 幸輝, 三浦 瑛絵, 深谷 猛, 岩下 武史, 中島 浩
The 2nd. cross-disciplinary Workshop on Computing Systems, Infrastructures, and Programming (xSIG2018), 30 May 2018, 情報処理学会 ARC/HPC/OS/PRO 各研究会, Oral presentation
東京都, Japan, [Domestic Conference] - Enhancement of Algebraic Block Multi-Color Ordering for ILU Preconditioning and Its Performance Evaluation in Preconditioned GMRES Solver
Senxi Li, Takeshi Iwashita, Takeshi Fukaya
The 2nd. cross-disciplinary Workshop on Computing Systems, Infrastructures, and Programming (xSIG2018), 30 May 2018, 情報処理学会 ARC/HPC/OS/PRO 各研究会, Oral presentation
東京都, Japan, [Domestic Conference] - Performance Evaluation of Time-Space Tiling with Tile-Level Parallelism for Iterative Stencil Computations
Takeshi Fukaya, Takeshi Iwashita
2018 Conference on Advanced Topics and Auto Tuning in High-Performance Scientific Computing (ATAT in HPSC 2018), 26 Mar. 2018, Oral presentation
Tainan, Taiwan, Province of China, [International presentation] - Oakforest-PACSにおける一般化固有値計算の性能解析と性能予測
星 健夫, 福本 智哉, 深谷 猛, 山本 有作
日本応用数理学会 2018年研究部会連合発表会, 16 Mar. 2018, 日本応用数理学会, Oral presentation
吹田市, Japan, [Domestic Conference] - 高性能計算入門：より高速な計算を目指して
深谷 猛
日本応用数理学会若手の会主催 応用数理 学生・若手研究者のための研究交流会, 14 Mar. 2018, 日本応用数理学会 若手の会, Public discourse
吹田市, [Domestic Conference] - 複数のデータ構造を用いた疎行列ベクトル積のキャッシュブロッキング手法の検討と評価
石田 幸輝, 三浦 瑛絵, 深谷 猛, 岩下 武史, 中島 浩
日本応用数理学会若手の会主催 応用数理 学生・若手研究者のための研究交流会, 14 Mar. 2018, 日本応用数理学会 若手の会, Oral presentation
吹田市, Japan, [Domestic Conference] - An Approach to Accelerating the SpMV Kernel by Exploiting Specific Sparse Structures
Takeshi Fukaya, Koki Ishida, Akie Miura, Takeshi Iwashita, Hiroshi. Nakashima
SIAM Conference on Parallel Processing for Scientific Computing (SIAM PP18), 10 Mar. 2018, SIAM, Oral presentation
Tokyo, Japan, [International presentation] - Shifted Cholesky QR for Computing the QR Factorization for Ill-conditioned Matrices
Yuka Yanagisawa, Takeshi Fukaya, Yuji Nakatsukasa, Yusaku Yamamoto, Ranseshan Kannan
SIAM Conference on Parallel Processing for Scientific Computing (SIAM PP18), 09 Mar. 2018, SIAM, Oral presentation
Tokyo, Japan, [International presentation], ［Internationally co-authored］ - Effect of Algebraic Block Multi-Color Ordering for Multi-Threaded ILU-GMRES Solver
Senxi Li, Takeshi Fukaya, Takeshi Iwashita
SIAM Conference on Parallel Processing for Scientific Computing (SIAM PP18), 08 Mar. 2018, SIAM, Poster presentation
Tokyo, Japan, [International presentation] - Performance Evaluation of Tiled 3D FDTD Solver on Recent Multicore Processors
Takeshi Iwashita, Takeshi Fukaya
SIAM Conference on Parallel Processing for Scientific Computing (SIAM PP18), 07 Mar. 2018, SIAM, Oral presentation
Tokyo, Japan, [International presentation] - Current status of EigenExa, high-performance parallel dense eigensolver
Toshiyuki Imamura, Yusuke Hirota, Takeshi Fukaya
2018 International Workshop on Eigenvalue Problems: Algorithms; Software and Applications, in Petascale Computing (EPASA2018), 06 Mar. 2018, Poster presentation
Tsukuba, Japan, [International presentation] - Analysis and prediction of the performance in generalized eigenvalue solvers on Oakforest-PACS
Takeo Hoshi, Tomoya Fukumoto, Takeshi Fukaya, Yusaku Yamamoto
2018 International Workshop on Eigenvalue Problems: Algorithms; Software and Applications, in Petascale Computing (EPASA2018), 06 Mar. 2018, Poster presentation
Tsukuba, Japan, [International presentation] - Overview of the EigenExa project, past, present and future
Toshiyuki Imamura, Yusuke Hirota, Takeshi Fukaya
2018 International Workshop on Eigenvalue Problems: Algorithms; Software and Applications, in Petascale Computing (EPASA2018), 06 Mar. 2018, Oral presentation
Tsukuba, Japan, [International presentation] - 並列計算機上での反復型ステンシル計算に対する効果的な時空間タイリングとその応用
深谷 猛
科研費基盤B課題「O(1億)コア環境におけるスケーラブルな数値計算ソフトウェアの理論と応用」ワークショップ, 23 Jan. 2018, Oral presentation
札幌市, Japan, [Domestic Conference] - 疎行列のステンシル構造の活用による疎行列ベクトル積の性能向上の調査
深谷 猛, 三浦 瑛絵, 岩下 武史
大学ICT推進協議会 2017年度年次大会（AXIES2017）, 13 Dec. 2017, 一般社団法人 大学ICT推進協議会, Oral presentation
広島市, Japan, [Domestic Conference] - A parallel solver for a linear system with a symmetric sparse matrix by one-dissection ordering
Tomoki Nakano, Mitsuo Yokokawa, Takeshi Fukaya, Yusaku Yamamoto
Workbench on Sustained Simulation Performance (WSSP), 10 Oct. 2017, Oral presentation
Stuttgart, Germany, [International presentation] - 時空間タイリングを用いた反復型ステンシル計算とその応用
岩下 武史, 深谷 猛
日本機械学会 第30回計算力学講演会（CMD2017）, 17 Sep. 2017, 日本機械学会, Oral presentation
東大阪市, Japan, [Domestic Conference] - TSQRアルゴリズムにおける三角行列のリダクション処理に関する考察
深谷 猛
日本応用数理学会2017年度年会, 06 Sep. 2017, 日本応用数理学会, Oral presentation
東京都, Japan, [Domestic Conference] - Temporal and spatial tiling technique with tile-level parallelism and its application to 3D FDTD method
Takeshi Fukaya
Sapporo Summer HPC Seminar 2017, 07 Aug. 2017, Information Initiative Center, Hokkaido University, Oral presentation
Sapporo, Japan, [International presentation] - 縦長行列のQR分解に対する通信削減型アルゴリズムの性能評価
深谷 猛, 山本 有作
第2回CDMSI（ポスト「京」重点課題（７））研究会, Jul. 2017, Poster presentation
11 Jul. 2017 - 12 Jul. 2017, 東京都, Japan, [Domestic Conference] - 複数の格納形式を利用した疎行列ベクトル積の高速化に関する検討
石田 幸輝, 三浦 瑛絵, 深谷 猛, 岩下 武史, 中島 浩
2017年ハイパフォーマンスコンピューティングと計算科学シンポジウム（HPCS2017）, 05 Jun. 2017, 情報処理学会 HPC研究会, Poster presentation
神戸市, Japan, [Domestic Conference] - H行列ベクトル積のスレッド並列化手法に関する性能評価
川村 卓人, 深谷 猛, 岩下 武史, 伊田 明弘
2017年ハイパフォーマンスコンピューティングと計算科学シンポジウム（HPCS2017）, 05 Jun. 2017, 情報処理学会 HPC研究会, Poster presentation
神戸市, Japan, [Domestic Conference] - Shifted CholeskyQR for Computing the factorization of ill-conditioned matrices
Yuka Yanagisawa, Takeshi Fukaya, Ramaseshan Kannan, Yuji Nakatsukasa, Yusaku Yamamoto, Oishi Shin’ichi
The International Workshop on Numerical Verification and its Applications 2017 (INVA2017), 16 Mar. 2017, Oral presentation
Miyakojima, Japan, [International presentation], ［Internationally co-authored］ - ステンシル構造を利用した疎行列ベクトル積の高速化に関する検討
三浦 瑛絵, 深谷 猛, 岩下 武史
日本応用数理学会 若手の会 第2回学生研究発表会, 05 Mar. 2017, 日本応用数理学会 若手の会, Poster presentation
東京都, Japan, [Domestic Conference] - Performance Evaluation of Time-Space Tiling Strategies for Iterative Stencil Computations on Multi/Many-Core CPU Systems
Takeshi Fukaya, Takeshi Iwashita
SIAM Conference on Computational Science and Engineering (CSE17), 28 Feb. 2017, SIAM, Oral presentation
Atlanta, United States, [International presentation] - 最近の計算機環境における基本的な行列計算カーネルの性能とその考察
深谷 猛
ワークショップ「行列計算のための数値計算法」, 20 Jan. 2017, Oral presentation
名古屋市, Japan, [Domestic Conference] - 時空間タイリングによる反復型ステンシル計算の性能向上に関する基礎評価
深谷 猛, 岩下 武史
大学ICT推進協議会 2016年度年次大会（AXIES2016）, 16 Dec. 2016, 一般社団法人 大学ICT推進協議会, Oral presentation
京都市, [Domestic Conference] - ScaLAPACKの性能分析と次世代アルゴリズム研究への指針
深谷 猛
計算物質科学における時空間アップスケーリングと数理手法, 29 Nov. 2016, Public discourse
調布市, Japan, [Domestic Conference] - マルチコア・メニーコア環境における反復型ステンシル計算と時空間タイリング
深谷 猛, 岩下 武史
日本応用数理学会2016年度年会, 07 Sep. 2016, 日本応用数理学会, Oral presentation
北九州市, Japan, [Domestic Conference] - Time-space tiling strategies for iterative stencil computations on multi/many-core CPU systems
Takeshi Fukaya, Takeshi Iwashita
Sapporo Summer HPC Seminar 2016, 22 Aug. 2016, Information Initiative Center, Hokkaido University, Oral presentation
Sapporo, Japan, [International presentation] - 反復型ステンシル計算のマルチコア・メニーコア向け実装に関する考察
深谷 猛, 岩下 武史
日本応用数理学会「行列・固有値問題の解法とその応用」研究部会 第21回研究会（SwoPP2016）, 09 Aug. 2016, 日本応用数理学会「行列・固有値問題の解法とその応用」研究部会, Oral presentation
松本市, Japan, [Domestic Conference] - Performance Evaluation of Verified Computation for Linear System on Supercomputer
Yusuke Morikura, Daichi Mukunoki, Takeshi Fukaya, Naoya Yamanaka
The 11th East Asia Section of SIAM Conference (EASIAM 2016), 20 Jun. 2016, EASIAM, Oral presentation
Macau, China, [International presentation] - 分散並列計算機における密行列ベクトル積の通信隠蔽実装の評価
川村 卓人, 深谷 猛, 岩下 武史
2016年ハイパフォーマンスコンピューティングと計算科学シンポジウム（HPCS2016）, 06 Jun. 2016, 情報処理学会 HPC研究会, Poster presentation
仙台市, Japan, [Domestic Conference] - An Impact of Tuning the Kernel of the Structured QR Factorization in the TSQR
Takeshi Fukaya, Toshiyuki Imamura
SIAM Conference on Parallel Processing for Scientific Computing (PP16), 14 Apr. 2016, SIAM, Oral presentation
Paris, France, [International presentation] - 分散並列計算環境における通信隠蔽手法を用いた密行列ベクトル積実装の性能評価
川村 卓人, 深谷 猛, 岩下 武史
日本応用数理学会 若手の会 第1回学生研究発表会, 03 Mar. 2016, 日本応用数理学会 若手の会, Poster presentation
神戸市, Japan, [Domestic Conference] - Performance evaluation of the tall-skinny QR factorization on recent parallel systems
Takeshi Fukaya, Yusaku Yamamoto, Toshiyuki Imamura
The 6th AICS International Symposium, 22 Feb. 2016, RIKEN AICS, Poster presentation
Kobe, [International presentation] - Performance Evaluation of Verified Computation for Linear Systems on Parallel Computers
Yusuke Morikura, Daichi Mukunoki, Takeshi Fukaya, Naoya Yamanaka, Shin’ichi Oishi
2nd Annual Meeting on Advanced Computing System and Infrastructure (ACSI 2016), 18 Jan. 2016, Poster presentation
福岡市, Japan, [Domestic Conference] - 線形計算アルゴリズムと通信回避
深谷 猛
研究会「数理構造保存を接点として数学・HPC・実科学のクロスオーバー」, 25 Nov. 2015, Oral presentation
東京都, Japan, [Domestic Conference] - Roundoff Error Analysis of the Choleskyqr2 and Related Algorithms
Yusaku Yamamoto, Yuji Nakatsukasa, Yuka Yanagisawa, Takeshi Fukaya
SIAM Conference on Applied Linear Algebra (LA15), 28 Oct. 2015, SIAM, Oral presentation
Atlanta, United States, [International presentation] - Performance Evaluation of the Choleskyqr2 Algorithm
Takeshi Fukaya, Yuji Nakatsukasa, Yuka Yanagiswa, Yusaku Yamamoto
SIAM Conference on Applied Linear Algebra (LA15), 27 Oct. 2015, SIAM, Oral presentation
Atlanta, United States, [International presentation] - The CholeskyQR2 algorithm and its applications
Takeshi Fukaya
20th ASE Seminar (Advanced Supercomputing Environment), 16 Oct. 2015, Information Technology Center, The University of Tokyo, Oral presentation
Tokyo, Japan, [International presentation] - Performance evaluation of the divide-and conquer method in the EigenExa eigensolver
Takeshi Fukaya, Toshiyuki Imamura
International Workshop on Eigenvalue Problems: Algorithms; Software and Applications, in Petascale Computing (EPASA2015), 15 Sep. 2015, Poster presentation
Tsukuba, Japan, [International presentation] - 重み付き内積空間における行列のQR分解アルゴリズムの考察－高性能計算の視点から
深谷 猛, 中務 佑治, 柳澤 優香, 山本 有作
日本応用数理学会2015年度年会, 09 Sep. 2015, 日本応用数理学会, Poster presentation
金沢市, Japan, [Domestic Conference] - CAHTR: Communication-Avoiding Householder TRidiagonalization
Toshiyuki Imamura, Takeshi Fukaya, Yusuke Hirota, Susumu Yamada, Masahiko Machida
International Conference on Parallel Computing (ParCo) 2015, 03 Sep. 2015, Oral presentation
Edinburgh, United Kingdom, [International presentation] - Moving a specified eigenvalue and eigenvector
Yuji Nakatsukasa, Takeshi Fukaya, Agnieszka Miedlar
The 8th International Congress on Industrial and Applied Mathematics (ICIAM2015), 10 Aug. 2015, ICIAM, Oral presentation
Beijing, China, [International presentation], ［Internationally co-authored］ - ペタ・ポストペタスケールシステムにおける密行列向けアルゴリズムの実行時間：EigenExaの開発を通して得られた実測データに基づく考察
深谷 猛, 山本 有作, 今村 俊幸
日本応用数理学会「行列・固有値問題の解法とその応用」研究部会 第19回研究会（SWoPP2015）, 05 Aug. 2015, 日本応用数理学会「行列・固有値問題の解法とその応用」研究部会, Oral presentation
別府市, Japan, [Domestic Conference] - コレスキーQR分解を用いたブロック直交変換の生成
深谷 猛, 中務 佑治, 山本 有作
第44回数値解析シンポジウム（NAS2015）, 09 Jun. 2015, Oral presentation
甲府市, Japan, [Domestic Conference] - ストペタスケール計算機上での密行列向け固有値ソルバーの性能の展望
深谷 猛, 山本 有作, 今村 俊幸
2015年ハイパフォーマンスコンピューティングと計算科学シンポジウム（HPCS2015）, 19 May 2015, 情報処理学会 HPC研究会, Poster presentation
東京都, [Domestic Conference] - Performance Evaluation of EigenExa Dense Eigensolver on the Oakleaf-Fx Supercomputer System
Takeshi Fukaya, Toshiyuki Imamura
SIAM Conference on Computational Science and Engineering (CSE15), 14 Mar. 2015, SIAM, Oral presentation
Salt Lake City, United States, [International presentation] - Numerical Eigenvalue Engine towards Extreme-scale Computing Era
Toshiyuki Imamura, Takeshi Fukaya, Yusuke Hirota, Susumu Yamada, Masahiko Machida
SIAM Conference on Computational Science and Engineering (CSE15), 14 Mar. 2015, SIAM, Oral presentation
Salt Lake City, United States, [International presentation] - オンライン自動チューニングのための性能モデルの構築法 ～ 正方行列の特異値分解を例にして ～
長島 聖児, 深谷 猛, 山本有作, 横川三津
日本応用数理学会2015年研究部会連合発表会, 06 Mar. 2015, 日本応用数理学会, Oral presentation
東京都, Japan, [Domestic Conference] - CholeskyQR2: an algorithm of the Cholesky QR factorization with reorthogonalization
Takeshi Fukaya
2015 Conference on Advanced Topics and Auto Tuning in High Performance Scientific Computing (2015 ATAT in HPSC), 28 Feb. 2015, Oral presentation
Taipei, Taiwan, Province of China, [International presentation] - Performance evaluation of the EigenExa eigensolver on the Oakleaf-FX supercomputing system
Takeshi Fukaya, Imamura Toshiyuki
Annual Meeting on Advanced Computing System and Infrastructure (ACSI 2015), 27 Jan. 2015, Oral presentation
つくば市, Japan, [Domestic Conference] - 高性能計算におけるコレスキーQR分解
深谷 猛
第12回計算数学研究会, 28 Dec. 2014, Invited oral presentation
焼津市, Japan, [Invited], [Domestic Conference] - Performance evaluation og the EigenExa dense eigensolver on the K computer
Takeshi Fukaya, Toshiyuki Imamura
5th AICS International Symposium, 08 Dec. 2014, RIKEN AICS, Poster presentation
Kobe, Japan, [International presentation] - Modeling the performance of parallel dense eigensolvers on peta/post-petascale systems
Takeshi Fukaya
JST/CREST International Symposium on Post Petascale System Software (ISP2S2), 02 Dec. 2014, Poster presentation
Kobe, Japan, [International presentation] - コレスキー分解に基づくQR分解の計算方法について
深谷 猛
第8回協定講座シンポジウム「計算科学 次代を担う若手の集い2014」, 11 Sep. 2014, 神戸大学大学院 システム情報学研究科, Poster presentation
神戸市, Japan, [Domestic Conference] - シフト付きコレスキーQR分解アルゴリズムの提案
柳澤 優香, 深谷 猛, 中務 佑治, Kannan Ramaseshan, 山本 有作, 大石 進一
日本応用数理学会2014年度年会, Sep. 2014, 日本応用数理学会, Oral presentation
03 Sep. 2014 - 05 Sep. 2014, 東京都, Japan, [Domestic Conference], ［Internationally co-authored］ - 大規模並列計算機上での再直交化付きコレスキーQR分解の性能評価
深谷 猛, 中務 佑治, 柳澤 優香, 山本 有作
本応用数理学会2014年度年会, Sep. 2014, 日本応用数理学会, Oral presentation
03 Sep. 2014 - 05 Sep. 2014, 東京都, Japan, [Domestic Conference] - ハウスホルダー変換のブロック化と通信回数削減に関する一考察
深谷 猛, 山本 有作, 今村 俊幸
日本応用数理学会「行列・固有値問題の解法とその応用」研究部会第17回研究会（SWoPP2014）, 28 Jul. 2014, 日本応用数理学会「行列・固有値問題の解法とその応用」研究部会, Oral presentation
新潟市, Japan, [Domestic Conference] - EigenExa: high performance dense eigensolver, present and future
Toshiyuki Imamura, Yusuke Hirota, Takeshi Fukaya, Susumu Yamada, Masahiko Machida
8th International Workshop on Parallel Matrix Algorithm and Applications (PMSS14), 2014, Oral presentation
02 Jul. 2014 - 04 Jul. 2014, Lugano, Switzerland, [International presentation] - 通信削減型QR分解アルゴリズムと自動チューニング
深谷 猛
第9回AT研究会オープンアカデミックセッション（ATOS9）, 12 May 2014, 自動チューニング研究会, Oral presentation
東京都, Japan, [Domestic Conference] - A Communication-Avoiding Algorithm for the Gram-Schmidt Orthogonalization
Takeshi Fukaya
2014 Conference on Advanced Topics and Auto Tuning in High Performance Scientific Computing (2014 ATAT in HPSC), Mar. 2014, Oral presentation
14 Mar. 2014 - 15 Mar. 2014, Taipei, Taiwan, Province of China, [International presentation] - Cholesky-QR and Householder-QR factorizations in nonstandard inner product spaces
Yuka Yanagisawa, Yuji Nakatsukasa, Takeshi Fukaya
International Workshop on Eigenvalue Problems: Algorithms; Software and Applications, in Petascale Computing (EPASA 2014), Mar. 2014, Poster presentation
07 Mar. 2014 - 09 Mar. 2014, Tsukuba, Japan, [International presentation] - An overview of parallel algorithms for tall-skinny QR factorizations
Takeshi Fukaya, Yusaku Yamamoto, Toshiyuki Imamura
International Workshop on Eigenvalue Problems: Algorithms; Software and Applications, in Petascale Computing (EPASA 2014), Mar. 2014, Poster presentation
07 Mar. 2014 - 09 Mar. 2014, Tsukuba, Japan, [International presentation] - Auto-tuning Tall and Skinny QR Factorization
Takeshi Fukaya, Yusaku Yamamoto
SIAM Conference on Parallel Processing for Scientific Computing (PP14), Feb. 2014, SIAM, Oral presentation
18 Feb. 2014 - 21 Feb. 2014, Portland, United States, [International presentation] - グラム・シュミットの直交化に基づくTSQRアルゴリズムとその性能評価
深谷 猛, 山本 有作, 今村 俊幸
日本応用数理学会「行列・固有値問題の解法とその応用」研究部会第16回研究会, 26 Dec. 2013, 日本応用数理学会「行列・固有値問題の解法とその応用」研究部会, Oral presentation
東京都, Japan, [Domestic Conference] - 大規模並列環境における縦長行列のQR分解の性能評価
深谷 猛, 山本 有作, 今村 俊幸
第11回計算数学研究会, Nov. 2013, Oral presentation
02 Nov. 2013 - 04 Nov. 2013, 三朝町, Japan, [Domestic Conference] - 超並列環境におけるTSQRアルゴリズムの性能に関する一考察
深谷 猛
第5回協定講座シンポジウム「計算科学 次代を担う若手の集い2013」, 30 Sep. 2013, 神戸大学大学院 システム情報学研究科, Poster presentation
神戸市, Japan, [Domestic Conference] - オンライン自動チューニング数理基盤ライブラリATMathCoreLibの特異値分解問題への適用
長島 聖児, 深谷 猛, 山本 有作
日本応用数理学会2013年度年会, Sep. 2013, 日本応用数理学会, Oral presentation
09 Sep. 2013 - 11 Sep. 2013, 福岡市, Japan, [Domestic Conference] - ブロックヤコビ法に基づく固有値解法の超並列計算機上での実装
工藤 周平, 高橋 佑輔, 深谷 猛, 山本 有作
日本応用数理学会2013年度年会, Sep. 2013, 日本応用数理学会, Oral presentation
09 Sep. 2013 - 11 Sep. 2013, 福岡市, Japan, [Domestic Conference] - 京コンピュータにおける対称密行列向け固有値計算プログラムの性能評価と性能予測
深谷 猛, 今村 俊幸, 山本 有作
日本応用数理学会2013年度年会, Sep. 2013, 日本応用数理学会, Oral presentation
09 Sep. 2013 - 11 Sep. 2013, 福岡市, Japan, [Domestic Conference] - 超並列環境における縦長行列のQR分解に対する種々の計算方法の性能比較
深谷 猛, 山本 有作
第42回数値解析シンポジウム（NAS2013）, Jun. 2013, Oral presentation
12 Jun. 2013 - 14 Jun. 2013, 松山市, Japan, [Domestic Conference] - 京における密行列固有値ソルバEigen-Kの性能評価と性能モデリング
深谷 猛, 今村 俊幸, 山本 有作
SACSIS2013 -先進的計算基盤システムシンポジウム, May 2013, Poster presentation
22 May 2013 - 24 May 2013, 仙台市, Japan, [Domestic Conference] - Performance Evaluation and Tuning of Tall Skinny Type QR Factorization on the K Computer
Takeshi Fukaya, Yusaku Yamamoto
2013 Conference on Advanced Topics and Auto Tuning in High Performance Scientific Computing (2013 ATAT in HPSC), Mar. 2013, Oral presentation
27 Mar. 2013 - 29 Mar. 2013, Taipei, Taiwan, Province of China, [International presentation] - Performance Modeling of the Eigen-K Dense Eigensolver on Massively Parallel Machines
Takeshi Fukaya, Toshiyuki Imamura and Yusaku Yamamoto
SIAM Conference on Computational Science and Engineering (CSE13), Feb. 2013, SIAM, Oral presentation
25 Feb. 2013 - 01 Mar. 2013, Boston, United States, [International presentation] - TSQRアルゴリズムに基づくQR分解の並列計算に対する自動チューニング
深谷 猛
日本応用数理学会 若手の会 単独研究集会, 26 Dec. 2012, 日本応用数理学会 若手の会, Invited oral presentation
東京都, Japan, [Invited], [Domestic Conference] - ハウスホルダーQR分解の数値計算アルゴリズムと高性能計算のための工夫
深谷 猛
一橋大学 第14回「数理科学セミナー」, 21 Nov. 2012, 一橋大学 商学研究科, Invited oral presentation
東京都, Japan, [Invited], [Domestic Conference] - 超並列環境向け固有値計算プログラムの性能予測モデルの開発
深谷 猛
E-サイエンス若手・女性研究者シンポジウム2012, 17 Oct. 2012, 東京大学情報基盤センター, Oral presentation
柏市, Japan, [Domestic Conference] - SMP上での並列QR分解に対する自動チューニングの検討
深谷 猛, 山本 有作, 張 紹良
日本応用数理学会2012年度年会, Aug. 2012, 日本応用数理学会, Oral presentation
28 Aug. 2012 - 02 Sep. 2012, 稚内市, Japan, [Domestic Conference] - QR分解の並列計算における自動チューニングの検討
深谷 猛, 山本 有作, 張 紹良
第2回協定講座シンポジウム「計算科学 次代を担う若手の集い」, 23 Aug. 2012, 神戸大学大学院 システム情報学研究科, Poster presentation
神戸市, Japan, [Domestic Conference] - TSQR アルゴリズムを用いたSMP 上でのQR 分解計算に対する自動チューニングの検討
深谷 猛, 山本 有作, 張 紹良
第41回数値解析シンポジウム（NAS2012）, Jun. 2012, Poster presentation
06 Jun. 2012 - 08 Jun. 2012, 渋川市, Japan, [Domestic Conference] - ブロックQR分解アルゴリズムの性能最適化 -動的計画法を利用したブロック分割方法の決定
深谷 猛, 山本 有作, 張 紹良
第1回協定講座シンポジウム「計算アルゴリズムと化学・生物学の融合」, 17 Feb. 2012, 神戸大学大学院 システム情報学研究科, Poster presentation
神戸市, Japan, [Domestic Conference] - Automatic Performance Tuning for the Blocked Householder QR Algorithm
Takeshi Fukaya, Yusaku Yamamoto, Shao-Liang Zhang
The 7th East Asia SIAM Conference & RIMS Workshop on Methods in Industrial and Applied Mathematics, Jun. 2011, EASIAM/RIMS, Oral presentation
27 Jun. 2011 - 29 Jun. 2011, Kiakyushu, Japan, [International presentation] - ブロックQR分解アルゴリズムの性能最適化 －ブロック化による性能向上についての考察－
深谷 猛, 山本 有作, 張 紹良
第40回数値解析シンポジウム（NAS2011）, Jun. 2011, Poster presentation
20 Jun. 2011 - 22 Jun. 2011, 鳥羽市, Japan, [Domestic Conference] - Auto-tuning for BLAS-based Matrix Computations
Takeshi Fukaya, Yusaku Yamamoto, Shao-Liang Zhang
SIAM Conference on Computational Science and Engineering (CSE11), Feb. 2011, SIAM, Oral presentation
28 Feb. 2011 - 04 Mar. 2011, Reno, United States, [International presentation] - 動的計画法に基づく密行列計算アルゴリズムの再帰的ブロック化
深谷 猛, 山本 有作, 張 紹良
2011年ハイパフォーマンスコンピューティングと計算科学シンポジウム（HPCS2011）, Jan. 2011, 情報処理学会 HPC研究会, Poster presentation
18 Jan. 2011 - 19 Jan. 2011, つくば市, Japan, [Domestic Conference] - 密行列計算の再帰構造を利用した適応的なブロック化
深谷 猛, 山本 有作, 張 紹良
2010年度特異値・固有値合同ワークショップ, 27 Nov. 2010, Oral presentation
つくば市, Japan, [Domestic Conference] - LU分解アルゴリズムにおけるブロック分割法と性能の関係について
深谷 猛, 山本 有作, 張 紹良
第８回計算数学研究会, Oct. 2010, Poster presentation
29 Oct. 2010 - 31 Oct. 2010, 神戸市, Japan, [Domestic Conference] - 動的計画法によるQR分解のブロック分割法の決定
深谷 猛, 山本 有作, 張 紹良
第39回数値解析シンポジウム（NAS2010）, May 2010, Poster presentation
26 May 2010 - 28 May 2010, 鳥羽市, Japan - A Dynamic Programming Approach to Auto-Tuning the Blocking Strategy For the Householder QR Decomposition
Takeshi Fukaya, Yusaku Yamamoto and Shao-Liang Zhang
Workshop on Advanced Auto-tuning on Numerical Software (AANS2010), 02 Apr. 2010, Oral presentation
Tokyo, Japan, [International presentation] - An Approach to Automatic Tuning for the Parallel Householder Qr Decomposition
Takeshi Fukaya, Yusaku Yamamoto, Shao-Liang Zhang
SIAM Conference on Parallel Processing for Scientific Computing (PP10), Feb. 2010, SIAM, Oral presentation
24 Feb. 2010 - 26 Feb. 2010, Seattle, United States, [International presentation] - A Dynamic Programming Approach to Performance Optimization for the QR Decomposition
Takeshi Fukaya, Yusaku Yamamoto, Shao-Liang Zhang
International Symposium of Electronic Structure Calculations, Dec. 2009, Poster presentation
07 Dec. 2009 - 09 Dec. 2009, Tokyo, Japan, [International presentation] - マルチコア環境向けハウスホルダーQR 分解アルゴリズムの性能チューニング
深谷 猛, 山本 有作, 張 紹良
特異値・固有値合同ワークショップ, Nov. 2009, Oral presentation
21 Nov. 2009 - 22 Nov. 2009, つくば市, Japan, [Domestic Conference] - ハウスホルダーQR分解の並列計算の効率化
深谷 猛, 山本 有作, 張 紹良
第7回計算数学研究会, Oct. 2009, Oral presentation
16 Oct. 2009 - 18 Oct. 2009, 北塩原村, Japan, [Domestic Conference] - An Approach to Automatic Tuning for Parallel Householder QR Decomposition
Takeshi Fukaya, Yusaku Yamamoto, Shao-Liang Zhang
The Fourth International Workshop on Automatic Performance Tuning (iWAPT 2009), Oct. 2009, Poster presentation
01 Oct. 2009 - 02 Oct. 2009, Tokyo, Japan, [International presentation] - Totally Nonnegative帯行列向けqd法へのシフト導入について
山本 有作, 深谷 猛
日本応用数理学会2009年度年会, Sep. 2009, 日本応用数理学会, Oral presentation
28 Sep. 2009 - 30 Sep. 2009, 大阪市, Japan, [Domestic Conference] - Totally Nonnegativeな帯行列に対するqd法
山本 有作, 深谷 猛
第38回数値解析シンポジウム (NAS2009), Jun. 2009, Poster presentation
15 Jun. 2009 - 17 Jun. 2009, 東伊豆町, Japan, [Domestic Conference] - A Dynamic Programming Approach to Optimizing the Blocking Strategy for the Householder QR Decomposition
Takeshi Fukaya, Yusaku Yamamoto, Shao-Liang Zhang
The 2nd International Conference in Mathematical Modelling and Computation and The 5th East Asia SIAM Conference, Jun. 2009, EASIAM, Oral presentation
08 Jun. 2009 - 10 Jun. 2009, Bandar Seri Begawan, Brunei Darussalam, [International presentation] - ハウスホルダーQR分解のためのブロック分割法の動的決定
深谷 猛, 山本 有作, 張 紹良
第6回計算数学研究会, Mar. 2009, Oral presentation
16 Mar. 2009 - 18 Mar. 2009, 熱海市, Japan, [Domestic Conference] - A Dynamic Programming Approach to Auto-Tuning the Blocking Strategy For the Householder QR Decomposition
Takeshi Fukaya, Yusaku Yamamoto, Shao-Liang Zhang
SIAM Conference on Computational Science and Engineering (CSE09), Mar. 2009, SIAM, Oral presentation
02 Mar. 2009 - 06 Mar. 2009, Miami, United States, [International presentation] - ハウスホルダーQR分解におけるブロック分割パターンの最適化
深谷 猛, 山本 有作, 張 紹良
日本応用数理学会「行列・固有値問題の解法とその応用」研究部会 第5回研究会（SWoPP2008）, 05 Aug. 2008, 日本応用数理学会「行列・固有値問題の解法とその応用」研究部会, Oral presentation
佐賀市, Japan, [Domestic Conference] - Level3 BLASを用いたQR分解アルゴリズムの性能評価
深谷 猛, 山本 有作
第5回計算数学研究会, Oct. 2007, Poster presentation
27 Oct. 2007 - 29 Oct. 2007, 新潟市, Japan, [Domestic Conference]
Affiliated academic society
Research Themes
- Next-generation high-performance linear solver for future computational science and engineering
Grants-in-Aid for Scientific Research
Apr. 2023 - Mar. 2027
岩下 武史, 塙 敏博, 伊田 明弘, 美舩 健, 横田 理央, 高橋 康人, 今倉 暁, 深谷 猛
Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (A), Hokkaido University, Coinvestigator, 23H00462 - society5.0におけるデータ解析に資する高性能線形計算技術の研究
科学研究費助成事業 基盤研究(C)
Apr. 2021 - Mar. 2024
深谷 猛, 相島 健助
本研究課題では，Society 5.0におけるビッグデータ解析の基盤となり得る新しい線形計算アルゴリズムの研究開発を行う．エッジコンピューティングに代表される，従来のスーパーコンピュータとは異なる特徴を有する分散並列計算インフラ上で，IoTなどから生成される分散データを効率的に解析するために必要となる線形計算技術に関して，HPCと数理の両方の知見に基づいてアルゴリズムの研究開発を実施する．
2021年度は，分散並列環境における代表的なデータ分析手法（例：主成分分析，回帰分析）の現状に関する調査を行った．調査の結果，組み込み機器におけるストリームデータの主成分分析等の具体的なアプリケーション事例の現状を把握することができた．今回の調査の限りでは，アルゴリズムや実装方法において，性能改善に向けた検討の余地が十分にあることが分かった．一方で，使用メモリ量など，従来のHPCアプリケーションとは評価尺度の優先度が異なることも確認できた．今後は，今回の調査結果を踏まえて，本課題で取り組む具体的な問題設定や評価尺度などを整理する．
上述の調査と並行して，これまで研究を行ってきた行列計算アルゴリズムの中で，本課題と関わりの深い，行列のQR分解のアルゴリズムに関する研究を実施した．具体的には，縦長行列のQR分解を行う様々なアルゴリズムに関して，異なる特徴を持ったスーパーコンピュータ上での実行時間を評価した．特に，全体の実行時間に加えて，内部の通信時間などに関する詳細な測定を実施しており，得られた結果を用いて各アルゴリズムの実行時間の性能モデルを構築することで，エッジコンピューティング環境における各アルゴリズムの実行コストの予測等が可能となる．
日本学術振興会, 基盤研究(C), 北海道大学, Principal investigator, Competitive research funding, 21K11909 - 低精度・低信頼性演算を活用した数値計算アルゴリズムの創出
戦略的創造研究推進事業 さきがけ
Nov. 2020 - Mar. 2024
深谷猛
国立研究開発法人 科学技術振興機構, Principal investigator, Competitive research funding, JPMJPR20M8 - Development of technology for scientific simulations only using integer arithmetic for next-generation computers
Grants-in-Aid for Scientific Research Grant-in-Aid for Challenging Research (Exploratory)
Jul. 2020 - Mar. 2023
岩下 武史, 深谷 猛
2021年度の研究実績の概要は，以下に示す通りである．① 代表的な反復型ステンシル計算であるFDTD（Finite Difference Time Domain）法について，整数演算（固定小数点演算）のみを用いて，解析を行う方法についてその基本的な実装方針を構築した．本方針では，解析対象となる物理空間を複数の部分領域に分割し，部分領域ごとに異なるスケーリングファクタを用いることで，各領域内の物理量を与えられたビット幅の整数（固定小数点数）で表現する方式を採用する．領域間での物理量の連続性を保つ方策が実装面では必要となる．② 次世代の計算デバイスにおいて，高性能な整数演算処理はSIMD型の整数演算命令として実装される可能性がある．そこで，線形反復法を対象として，その代表的な前処理手法であるILU分解前処理をSIMD演算を前提として高速化する方法を考案し，性能評価を行った．本研究成果について口頭発表を行うとともに，学術論文としての発表準備を進めた．③ 2020年度に考案した整数演算のみを使用した線形ソルバは反復改良法の概念を利用しており，広義には混合精度演算を用いたソルバの一種と理解できる．実際，2021年度に発表された英国マンチェスター大学の数値線形代数における混合精度演算技術のレビュー論文において，本研究の成果が引用されている．そこで，反復改良法に基づく混合精度演算を利用した線形ソルバの性能評価について，主に計算結果の精度面から評価を行った．
Japan Society for the Promotion of Science, Grant-in-Aid for Challenging Research (Exploratory), Hokkaido University, Coinvestigator, Competitive research funding, 20K21782 - High performance linear solver for advanced computational electromagnetics
Grants-in-Aid for Scientific Research Grant-in-Aid for Scientific Research (B)
Apr. 2019 - Mar. 2022
IWASHITA TAKESHI
In this study, we studied the acceleration of the linear solver to make the electromagnetic field analysis based on the finite element and the boundary element methods more efficiently. Research was conducted from both the computer science and the mathematical approaches. From the computer science side, we studied the parallel in time technique, the preconditioning techniques in which SIMD instructions or accelerators are effectively used. We also studied the mixed precision computing in the context of iterative linear solvers. From the mathematical side, we focused on the situation in which a sequence of linear systems was solved. We proposed a technique based on error vector sampling and confirmed its effectiveness. In these studies, we attained many research results which were reported in journals and conferences.
Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (B), Hokkaido University, Coinvestigator, Competitive research funding, 19H04122 - Study on improving algorithms for tensor decomposition based on the HPC viewpoint
Grants-in-Aid for Scientific Research Grant-in-Aid for Early-Career Scientists
Apr. 2018 - Mar. 2021
Fukaya Takeshi
In this research, we aimed for improving algorithms of tensor decomposition, which is one of building blocks in data science applications. In addition to the traditional mathematical viewpoint, we investigated efficient algorithms based on the HPC viewpoint, in which the characteristics of recent computers are considered. Through this research, we found that a dominant computation in a typical tensor decomposition can be accelerated by appropriately selecting kernels depending on the conditions such as the size of a tensor. We also presented a new efficient algorithm for a matrix computation deeply related to tensor decomposition.
Japan Society for the Promotion of Science, Grant-in-Aid for Early-Career Scientists, Hokkaido University, Principal investigator, Competitive research funding, 18K18058 - Theory and Application of Scalable Numerical Software on an O(100M) core environment
Grants-in-Aid for Scientific Research Grant-in-Aid for Scientific Research (B)
Apr. 2015 - Mar. 2018
IMAMURA Toshiyuki, YAMAMOTO Yusaku, Todo Shinji
This research project aims to realize high performance numerical services investigated in the past based on new mathematical principles in the emerging computing system where tens of thousands to hundreds of millions of processing cores are installed. Giving two important themes, `Mixed-granularity numerical kernel' and `Asynchronous numerical algorithm,' we conducted; i) the research on the theory of asynchronous numerical algorithms. Also avoidance of communication and synchronization at a practical level, then CAHTR and a new method for the FDTD scheme were proposed. Furthermore, we have practiced; ii) promoting research on core numerical infrastructure technologies such as automatic tuning for scalable, lightweight code generation at super-many-core, and promoting innovative research leading to the next generation numerical calculation software.
Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (B), Institute of Physical and Chemical Research, Coinvestigator, Competitive research funding, 15H02709 - Fundamental research for improving the practicality of communication-avoiding matrix factorization algorithms
Grants-in-Aid for Scientific Research Grant-in-Aid for Young Scientists (B)
Apr. 2015 - Mar. 2018
Fukaya Takeshi
In order to reduce the communication cost in large-scale parallel computations, so-called communication-avoiding (CA) algorithms for matrix factorization have been actively studied. In this research, we aimed for improving the practicality of CA algorithms in realistic situations. We investigated fundamental techniques for CA algorithms: for example, those for implementing and tuning CA algorithms. We also compared different CA algorithms. In addition, we studied the performance modeling of CA algorithms.
Japan Society for the Promotion of Science, Grant-in-Aid for Young Scientists (B), Hokkaido University, Principal investigator, Competitive research funding, 15K16000 - 大規模行列計算のための階層的自動チューニング手法の開発
Grants-in-Aid for Scientific Research Grant-in-Aid for JSPS Fellows
2010 - 2011
深谷 猛
計算機環境の複雑化・多様化により,それぞれの条件(問題や計算機環境)に応じてアルゴリズムをチューニングすることが,高性能計算を実現するために不可避となっている。その際,従来の人手によるチューニングだけでなく,何らかの仕組みに基づいて計算機自身がチューニングを行う「自動チューニング」技術の開発が求められている。このような背景の下で,昨年度は基本的な行列計算の一つであるQR分解におけるブロック化の方法を自動的に決定する仕組みを構築し,有効な自動チューニング手法として期待できることを示した。そこで,本年度はこの手法をベースにして,実用化の観点から研究を進めた。
構築した手法では,動的計画法を用いることでアルゴリズムの候補を効率的に比較することが可能となっていた。また,比較の際に使用する評価値は性能予測モデルにより算出されることを前提としていた。そこで,本年度は,使用する性能予測モデルによる,チューニングの効果と実行コストの変化について考察した。また,行列サイズが大規模になった場合,全ての候補を比較することが困難になることが予想されるため,候補を限定してチューニングを行う手法の効果について検討した。さらに,限定の仕方を徐々に変化させることで,チューニングの効果とコストのトレードオブを効率的に制御する手法についても検討した。一方,並列計算を想定して,共有メモリ型並列計算機を用いてQR分解を行う場合の自動チューニング手法に関して検討した。並列計算では,TSQRと呼ばれるブロック分割が可能となり,同時に有効であることが知られているので,これを新たに取り入れたチューニング手法を構築し,その効果を検証した。
その他,QR分解以外として,LU分解アルゴリズムに対する自動チューニング手法を検討した。
以上の研究により,大規模行列計算アルゴリズムに対する実用的な自動チューニング手法の開発に向けた一つの方向性を示すとともに,その過程で解決すべき課題を具体的に明らかにすることができた。
Japan Society for the Promotion of Science, Grant-in-Aid for JSPS Fellows, Nagoya University, Principal investigator, Competitive research funding, 10J08599
Educational Organization
- Master's degree program, Graduate School of Information Science and Technology
- Doctoral (PhD) degree program, Graduate School of Information Science and Technology
