Organization
Information Technology Center High Performance Computing division Designated assistant professor
Title
Designated assistant professor
Contact information
External link
MUKUNOKI Daichi
Degree 1
-
博士（工学） （ 2013.11 筑波大学 ）
Research Interests 7
-
High performance computing
-
Accurate computation
-
Auto-tuning
-
Numerical computation
-
Reproducible computation
-
Parallel Computing
-
GPU computing
Research Areas 2
-
Informatics / High performance computing
-
Informatics / Computer system
Research History 13
-
Nagoya University Information Technology Center Designated assistant professor
2024.12
Country：Japan
-
Shibaura Institute of Technology Temporary Technical Staff
2024.4 - 2024.11
Country：Japan
-
Sony Interactive Entertainment Inc. Sr. Software Engineer
2023.11 - 2024.2
Country：Japan
-
RIKEN Center for Computational Science Large-scale Parallel Numerical Computing Technology Research Team Researcher
2019.4 - 2023.10
Country：Japan
-
Tokyo Woman's Christian University Graduate School of Science Postdoctoral Research Fellow
2017.10 - 2019.3
Country：Japan
-
RIKEN Advanced Institute of Computational Science Architecture Development Team, Flagship 2020 Project Postdoctoral Researcher
2017.4 - 2017.9
Country：Japan
-
RIKEN Advanced Institute of Computational Science Co-design Team, Flagship 2020 Project Postdoctoral Researcher
2015.5 - 2017.3
Country：Japan
-
RIKEN Advanced Institute for Computational Science Large-scale Parallel Numerical Computing Technology Research Team, Research Division Postdoctoral Researcher
2014.6 - 2017.9
Country：Japan
-
Japan Society for the Promotion of Science Research Fellow (PD)
2013.12 - 2014.5
Country：Japan
-
Japan Society for the Promotion of Science Research Fellow (DC2)
2013.4 - 2013.11
Country：Japan
-
Information Technology Center, The University of Tokyo Visiting Researcher
2021.11 - 2023.3
Country：Japan
-
RIKEN Center for Computational Science Architecture Development Team, Flagship 2020 Project Visiting Researcher
2017.10 - 2019.3
Country：Japan
-
RIKEN Advanced Institute of Computational Science Large-scale Parallel Numerical Computing Technology Research Team, Research Division Visiting Researcher
2017.10 - 2019.3
Country：Japan
Education 4
-
University of Tsukuba Graduate School of Systems and Information Engineering
2011.4 - 2013.11
Country： Japan
-
University of Tsukuba Graduate School of Systems and Information Engineering
2009.4 - 2011.3
Country： Japan
-
University of Tsukuba School of Library and Information Science
2006.4 - 2009.3
Country： Japan
-
Gifu National College of Technology
2001.4 - 2006.3
Country： Japan
Professional Memberships 2
-
Information Processing Society of Japan
2008
-
Auto-Tuning Resarch Group
Committee Memberships 34
-
The 15th International Conference on Parallel Processing & Applied Mathematics (PPAM 2024) Program Committee Member
2024
-
Mini Symposium: Exploring Arithmetic and Data Representation Beyond the Standard in HPC (at ICIAM 2023) Mini-Symposium Organizer
2023
-
2023 IEEE 16th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC-2023) Program Committee Member
2023
-
Special Session: Performance Optimization and Auto-Tuning of Software on Multicore/Manycore Systems (POAT 2023) (in conjunction with MCSoC-2023) Program Chair
2023
Committee type：Academic society
-
The 24th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2023) (in conjunction with IPDPS 2023) Program Committee Member
2023
-
The 22nd International Conference on Computational Science (ICCS 2022) Program Committee Member
2022
-
36th IEEE International Parallel & Distributed Processing Symposium (IPDPS 2022) Program Committee Member (Algorithm track)
2022
-
The International Conference on High Performance Computing in Asia-Pacific Region (HPC Asia 2022) Publicity Chair
2022
-
自動チューニング研究会 幹事（交流促進委員会）
2021 - 2023
Committee type：Academic society
-
情報処理学会論文誌コンピューティングシステム 編集委員
2020 - 2024
Committee type：Academic society
-
The International Conference for High Performance Computing, Networking, Storage, and Analysis (SC20) Research Poster Committee Member
2020
-
The 4th International Workshop on GPU Computing and AI (GCA'19) (in conjunction with CANDAR'19) Program Committee Member
2019
-
The Fourteenth International Workshop on Automatic Performance Tuning (iWAPT2019) (in conjunction with IPDPS 2019) Program Committee Member
2019
-
The 14th International Conference on Parallel Processing & Applied Mathematics (PPAM 2022) Program Committee Member
2022
-
Special Session: Auto-Tuning for Multicore and GPU (ATMG2022) (in conjunction with MCSoC-2022) Program Chair
2022
-
IEEE 22nd International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2021) (in conjunction with IPDPS 2021) Program Committee Member
2021
-
Workshop on Large-scale Parallel Numerical Computing Technology (LSPANC 2020 January) Program Committee Member
2020
-
The 21st IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2020) (in conjunction with IPDPS 2020) Program Committee Member
2020
-
2019 IEEE 13th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC-2019) Program Committee Member
2019
-
The 20th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2019) (in conjunction with IPDPS 2019) Program Committee Member
2019
-
Mini Symposium: Development of Numerical Computing Software on Emerging Computing Platforms (at SIAM PP 18) Mini-Symposium Organizer
2018
-
Special Session: Auto-Tuning for Multicore and GPU (ATMG 2018) (in conjunction with MCSoC-2018) Program Committee Member
2018
-
The Thirteenth International Workshop on Automatic Performance Tuning (iWAPT2018) (in conjunction with IPDPS 2018) Program Committee Member
2018
-
2018 IEEE 12th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC-2018) Program Committee Member
2018
-
The 19th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2018) (in conjunction with IPDPS 2018) Program Committee Member
2018
-
The Third International Workshop on GPU Computing and AI (GCA'18) (in conjunction with CANDAR'18) Program Committee Member
2018
-
The Second International Workshop on GPU Computing and AI (GCA'17) (in conjunction with CANDAR'17) Program Committee Member
2017
-
The 18th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2017) (in conjunction with IPDPS 2017) Program Committee Member
2017
-
The Twelfth International Workshop on Automatic Performance Tuning (iWAPT2017) (in conjunction with IPDPS 2017) Program Committee Member
2017
-
Special Session: Auto-Tuning for Multicore and GPU (ATMG 2017) (in conjunction with MCSoC-17) Program Committee Member
2017
-
The First International Workshop on GPU Computing and Applications (GCA'16) (in conjunction with CANDAR'16) Program Committee Member
2016
-
The 17th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2016) (in conjunction with IPDPS 2016) Program Committee Member
2016
-
The 16th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2015) (in conjunction with IPDPS 2015) Program Committee Member
2015
-
The 15th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2014) (in conjunction with IPDPS 2014) Program Committee Member
2014
Awards 9
-
Best Paper Award
2023.12 6th IEEE International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC 2023) Sparse Matrix-Vector Multiplication with Reduced-Precision Memory Accessor
Daichi Mukunoki, Masatoshi Kawai, Toshiyuki Imamura
-
Research Poster Award 2nd Place Winner
2022.6 ISC High Performance 2022 A Fast Infinite Precision Inner Product using Ozaki Scheme and Dot2, and Its Application to Reproducible Conjugate Gradient Solvers
Daichi Mukunoki, Katsuhisa Ozaki, Takeshi Ogita, Toshiyuki Imamura
-
RIKEN Ohbu Award 2021
2022.3
-
Research Poster Award
2021.6 ISC High Performance 2021 Accurate Matrix Multiplication on Binary128 using Ozaki Scheme
Daichi Mukunoki, Katsuhisa Ozaki, Takeshi Ogita, Toshiyuki Imamura
-
Best Research Poster Award
2019.9 Russian Supercomputing Days Accurate and Reproducible Linear Algebra Operations for Many-core Architectures
Daichi Mukunoki, Takeshi Ogita, Katsuhisa Ozaki
-
PRACE-ISC Research Poster Award 2017
2017.6 ISC High Performance 2017 Implementation & Evaluation of 2.5D Matrix Multiplication on K Computer
Daichi Mukunoki, Toshiyuki Imamura
-
IPSJ Yamashita SIG Research Award
2016 Information Processing Society of Japan
-
IPSJ Computer Science Research Award for Young Scientists
2013 Information Processing Society of Japan
-
Young Researcher Award
2013 IPSJ Special Interest Group on System Architecture
Papers 54
-
Katsuhisa Ozaki, Daichi Mukunoki, Takeshi Ogita
Japan Journal of Industrial and Applied Mathematics 2024.10
-
Stef Graillat, Fabienne Jézéquel, Theo Mary, Roméo Molina, Daichi Mukunoki
Lecture Notes in Computer Science Vol. 14803 page： 17 - 30 2024.8
-
Mixed-precision conjugate gradient algorithm using the groupwise update strategy Reviewed
Kensuke Aihara, Katsuhisa Ozaki, Daichi Mukunoki
Japan Journal of Industrial and Applied Mathematics 2024.2
-
Sparse Matrix-Vector Multiplication with Reduced-Precision Memory Accessor Reviewed
Daichi Mukunoki, Masatoshi Kawai, Toshiyuki Imamura
2023 IEEE 16th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC) page： 608 - 615 2023.12
-
Daichi Mukunoki, Katsuhisa Ozaki, Takeshi Ogita, Toshiyuki Imamura
Parallel Processing and Applied Mathematics page： 40 - 54 2023.4
-
Task Scheduling Strategies for Batched Basic Linear Algebra Subprograms on Many-core CPUs Reviewed
Daichi Mukunoki, Yusuke Hirota, Toshiyuki Imamura
Proc. 2021 IEEE 14th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC) page： 234 - 241 2021.12
-
A Rapid Euclidean Norm Calculation Algorithm that Reduces Overflow and Underflow. Reviewed
Takeyuki Harayama, Shuhei Kudo, Daichi Mukunoki, Toshiyuki Imamura, Daisuke Takahashi
Proc. The 2021 International Conference on Computational Science and Its Applications (ICCSA 2021), Lecture Notes in Computer Science Vol. 12949 page： 95 - 110 2021.9
-
Accurate Matrix Multiplication on Binary128 Format Accelerated by Ozaki Scheme Reviewed
Daichi Mukunoki, Katsuhisa Ozaki, Takeshi Ogita, Toshiyuki Imamura
Proc. The 50th International Conference on Parallel Processing (ICPP-2021) 2021.8
-
Matrix Engines for High Performance Computing: A Paragon of Performance or Grasping at Straws? Reviewed
Jens Domke, Emil Vatai, Aleksandr Drozd, Peng Chen, Yosuke Oyama, Lingqi Zhang 0001, Shweta Salaria, Daichi Mukunoki, Artur Podobas, Mohamed Wahib, Satoshi Matsuoka
Proc. 35th IEEE International Parallel & Distributed Processing Symposium (IPDPS 2021) page： 1056 - 1065 2021.6
-
Daichi Mukunoki, Katsuhisa Ozaki, Takeshi Ogita, Roman Iakymchuk
Proc. The International Conference on High Performance Computing in Asia-Pacific Region (HPCAsia 2021) page： 100 - 109 2021.1
-
Can We Avoid Rounding-Error Estimation in HPC Codes and Still Get Trustworthy Results? Reviewed
Fabienne Jézéquel, Stef Graillat, Daichi Mukunoki, Toshiyuki Imamura, Roman Iakymchuk
Proc. 13th International Workshop on Numerical Software Verification 2020 (NSV 20), Lecture Notes in Computer Science Vol. 12549 page： 163 - 177 2020.12
-
Performance and energy consumption of accurate and mixed-precision linear algebra kernels on GPUs. Reviewed
Daichi Mukunoki, Takeshi Ogita
J. Comput. Appl. Math. Vol. 372 page： 112701 - 112701 2020.7
-
DGEMM Using Tensor Cores, and Its Accurate and Reproducible Versions Reviewed
Daichi Mukunoki, Katsuhisa Ozaki, Takeshi Ogita, Toshiyuki Imamura
Proc. ISC High Performance 2020, Lecture Notes in Computer Science Vol. 12151 page： 230 - 248 2020.6
-
Design of an FPGA-Based Matrix Multiplier with Task Parallelism. Reviewed
Yiyu Tan, Toshiyuki Imamura, Daichi Mukunoki
Proc. International Conference on Parallel Computing (ParCo2019), Parallel Computing: Technology Trends Vol. 36 page： 241 - 250 2019
-
Reproducible BLAS Routines with Tunable Accuracy Using Ozaki Scheme for Many-Core Architectures. Reviewed
Daichi Mukunoki, Takeshi Ogita, Katsuhisa Ozaki
Proc. 13th International Conference on Parallel Processing and Applied Mathematics (PPAM2019), Lecture Notes in Computer Science Vol. 12043 page： 516 - 527 2019
-
Performance Analysis of 2D-compatible 2.5D-PDGEMM on Knights Landing Cluster. Reviewed
Daichi Mukunoki, Toshiyuki Imamura
Proc. International Conference on Computational Science (ICCS 2018), Lecture Notes in Computer Science Vol. 10862 page： 853 - 858 2018.6
-
Toshiyuki Imamura, Daichi Mukunoki, Yusuke Hirota, Susumu Yamada, Masahiko Machida
Proc. International Conference on Parallel Computing (ParCo2017), Advances in Parallel Computing page： 97 - 106 2017.9
-
Implementation and Performance Analysis of 2.5D-PDGEMM on the K Computer. Reviewed
Daichi Mukunoki, Toshiyuki Imamura
Proc. 12th International Conference on Parallel Processing and Applied Mathematics (PPAM2017), Lecture Notes in Computer Science Vol. 10777 page： 348 - 358 2017
-
Automatic Thread-Block Size Adjustment for Memory-Bound BLAS Kernels on GPUs. Reviewed
Daichi Mukunoki, Toshiyuki Imamura, Daisuke Takahashi
Proc. IEEE 10th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC-16) page： 377 - 384 2016
-
Daichi Mukunoki, Toshiyuki Imamura
Proc. IEEE International Conference on Cluster Computing (Cluster 2016) page： 144 - 145 2016
-
Fast Implementation of General Matrix-Vector Multiplication (GEMV) on Kepler GPUs. Reviewed
Daichi Mukunoki, Toshiyuki Imamura, Daisuke Takahashi
Proc. 23rd Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP 2015) page： 642 - 650 2015
-
Implementation and Evaluation of Triple and Quadruple Precision Floating-point Operations on GPUs Reviewed
Vol. 6 ( 1 ) page： 66 - 77 2013.1
-
Daichi Mukunoki, Daisuke Takahashi
Proc. 13th International Conference on Computational Science and Its Applications (ICCSA 2013), Part V, Lecture Notes in Computer Science Vol. 7975 page： 211 - 223 2013
-
Using Quadruple Precision Arithmetic to Accelerate Krylov Subspace Methods on GPUs. Reviewed
Daichi Mukunoki, Daisuke Takahashi
Proc. 10th International Conference on Parallel Processing and Applied Mathematics (PPAM 2013), Part I, Workshop on Numerical Algorithms on Hybrid Architectures, Lecture Notes in Computer Science Vol. 8384 page： 632 - 642 2013
-
Implementation and Evaluation of Triple Precision BLAS Subroutines on GPUs. Reviewed
Daichi Mukunoki, Daisuke Takahashi
Proc. 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops & PhD Forum (IPDPSW 2012), The 13th Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC-12) page： 1378 - 1386 2012
-
Implementation and Evaluation of Quadruple and Octuple Precision BLAS on GPUs Reviewed
Vol. 2011 ( 2011 ) page： 148 - 156 2011.1
-
Implementation and Evaluation of Quadruple Precision BLAS Functions on GPUs. Reviewed
Daichi Mukunoki, Daisuke Takahashi
Proc. 10th International Conference on Applied Parallel and Scientific Computing (PARA 2010), Part I, Lecture Notes in Computer Science Vol. 7133 page： 249 - 259 2010
-
Performance Evaluation of Adaptive-Precision SpMV with Reduced-Precision Formats
Stef Grailla, Fabienne Jézéquel, Théo Mary, Roméo Molina, Daichi Mukunoki
HAL Vol. hal-04261073 2023.10
-
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
CoRR Vol. abs/2004.04628 2020.4
-
GPUの単精度演算・Tensorコアを用いた行列積のエラーフリー変換
尾崎克久, 椋木大地, 荻田武史
日本応用数理学会年会講演予稿集(CD-ROM) Vol. 2020 2020
-
尾崎スキームを用いたbinary128による4倍精度行列積
椋木大地, 尾崎克久, 荻田武史
日本応用数理学会年会講演予稿集(CD-ROM) Vol. 2020 2020
-
オーバー・アンダーフローを抑えた高精度かつ高速な2ノルム計算手法
原山赳幸, 工藤周平, 椋木大地, 今村俊幸, 高橋大介
情報処理学会研究報告(Web) Vol. 2020 ( HPC-177 ) 2020
-
尾崎スキームによる高精度かつ再現性のあるBLAS実装
椋木大地, 荻田武史, 尾崎克久, 今村俊幸
日本応用数理学会年会講演予稿集(CD-ROM) Vol. 2019 2019
-
Level-3BLASに基づく高精度行列積計算法による高精度かつ再現性のあるBLASルーチンの実装とその最適化
椋木大地, 荻田武史, 尾崎克久
情報処理学会研究報告(Web) Vol. 2018 ( HPC-166 ) 2018
-
京コンピュータにおける2.5次元アルゴリズムを用いた分散並列行列積の実装と評価
椋木大地, 今村俊幸
情報処理学会研究報告(Web) Vol. 2017 ( HPC-159 ) 2017
-
KMATHLIB-High Performance and Scalable Numerical Library for the K Computer-
大井祥栄, 廣田悠輔, 椋木大地, 今村俊幸
日本応用数理学会年会講演予稿集(CD-ROM) Vol. 2016 2016
-
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(Web) Vol. 2016 ( HPC-157 ) 2016
-
コンシューマレンジのGPUに最適化した固有値ソルバーの実装と評価
今村俊幸, 椋木大地
情報処理学会研究報告(Web) Vol. 2016 ( HPC-157 ) 2016
-
CUDA-BLAS等の選択による最速GPU固有値ソルバーの性能評価
今村俊幸, 今村俊幸, 椋木大地, 山田進, 山田進, 町田昌彦, 町田昌彦
情報処理学会研究報告(Web) Vol. 2015 ( HPC-148 ) 2015
-
FFTを使った時間発展問題における累積誤差
佐々成正, 山田進, 町田昌彦, 椋木大地, 今村俊幸
日本応用数理学会年会講演予稿集(CD-ROM) Vol. 2015 2015
-
短尺浮動小数点形式の検討
椋木大地, 今村俊幸
情報処理学会研究報告(Web) Vol. 2015 ( HPC-152 ) 2015
-
京・FX10における倍々精度演算の高速化
佐々木信一, 菱沼利彰, 藤井昭宏, 田中輝雄, 椋木大地, 今村俊幸
情報処理学会研究報告(Web) Vol. 2015 ( HPC-151 ) 2015
-
SYMV・GEMVルーチン群のマルチGPU化とその評価
今村俊幸, 今村俊幸, 椋木大地, 山田進, 山田進, 町田昌彦, 町田昌彦
情報処理学会研究報告(Web) Vol. 2015 ( HPC-151 ) 2015
-
NVIDIA GPUにおけるメモリ律速なBLASカーネルのスレッド数自動選択手法
椋木大地, 今村俊幸, 高橋大介
情報処理学会研究報告(Web) Vol. 2015 ( HPC-150 ) 2015
-
NVIDIA GPUにおけるGEMVカーネルの自動チューニング
椋木大地, 今村俊幸, 高橋大介
計算工学講演会論文集(CD-ROM) Vol. 20 2015
-
CUDA-xSYMVの実装と評価
今村俊幸, 今村俊幸, 椋木大地, 山田進, 山田進, 町田昌彦, 町田昌彦
情報処理学会研究報告(Web) Vol. 2014 ( HPC-146 ) 2014
-
MaxwellアーキテクチャGPUにおける疑似倍精度演算を用いたDGEMMの実装と評価
椋木大地, 今村俊幸
情報処理学会研究報告(Web) Vol. 2014 ( ARC-213 ) 2014
-
椋木大地, 高橋大介
研究報告ハイパフォーマンスコンピューティング（HPC） Vol. 2013 ( 5 ) page： 1 - 7 2013.2
-
GPUにおける4倍精度浮動小数点演算を用いたクリロフ部分空間法の高速化
椋木大地, 椋木大地, 高橋大介
情報処理学会研究報告(Web) Vol. 2013 ( HPC-140 ) 2013
-
GPUにおける4倍精度演算を用いた疎行列反復解法の実装と評価
椋木大地, 高橋大介
情報処理学会研究報告(CD-ROM) Vol. 2012 ( 5 ) 2013
-
GPUにおける4倍精度演算を用いた疎行列反復解法の実装と評価
椋木大地, 高橋大介
研究報告ハイパフォーマンスコンピューティング（HPC） Vol. 2012 ( 37 ) page： 1 - 8 2012.12
-
GPUによる3倍精度浮動小数点演算の検討
椋木大地, 高橋大介
情報処理学会研究報告(CD-ROM) Vol. 2011 ( 4 ) 2011
-
GPUによる4倍精度BLASの実装と評価
椋木大地, 高橋大介
計算工学講演会論文集 Vol. 15 ( 2 ) 2010
-
Implementation and Evaluation of Quadruple Precision BLAS on GPU
椋木大地, 高橋大介
情報処理学会研究報告(CD-ROM) Vol. 2009 ( 4 ) 2009
KAKENHI (Grants-in-Aid for Scientific Research) 6
-
Development of Accurate and Validated Matrix Computation Software for Next Generation Supercomputers
Grant number：20KK0259 2022.4 - 2023.10
Japan Society for the Promotion of Science (JSPS) Fund for the Promotion of Joint International Research (Fostering Joint International Research (A)) Fund for the Promotion of Joint International Research (Fostering Joint International Research (A))
Daichi Mukunoki
Authorship：Principal investigator
Grant amount：\9230000 （ Direct Cost: \7100000 、 Indirect Cost：\2130000 ）
-
超並列計算環境のための高精度かつ再現性のある行列計算ライブラリの開発
Grant number：19K20286 2019.4 - 2022.3
日本学術振興会 科学研究費助成事業 若手研究 若手研究
椋木 大地
本研究の目的は数値計算における計算の高精度化および再現性の保証を実現し，かつ最先端の超並列計算機アーキテクチャにおいて高性能を実現できるBLASライブラリの開発を行うことである．本研究では4つの手法：(1)尾崎スキーム，(2)ExBLASスキーム，(3)DotKスキーム，(4)CADNAスキームに着目し，このうち(1)を主たる手法として検討する．
2019年度は主として(1)(4)に関する進捗が得られた．(1)に関してはCPU・GPU向けのBLASの基本ルーチンを開発し，オープンソースソフトウェアとして公開した．またこれらに関する査読付き論文を国際学会（PPAM2019）において発表した．さらにその応用として，疎行列反復解法（CG法）への適用，FP16の活用に関する研究を前倒しして実施した（これらは当初2021年度の実施を予定していた）．このうち後者については，FP16/32の混合精度ハードウェアであるTensor Coresを活用して高速に高精度・再現性のある実装を行う方法を開発し，査読付き論文が国際学会（ISC2020）に採択された．また(4)CADNAスキームについては，その開発元であり共同研究を進めているソルボンヌ大学側で新しい手法が考案され，共著者として参加した論文を国際学会に投稿した（プレプリント公開済み，現在査読中）．
一方，計算結果の精度を担保しながら数値計算に用いられる演算精度を最適化して計算の高速化，省電力化を実現する方法の研究を開始した．本科研費課題で取り組む上記(1)-(4)の手法はその要素技術となりうるため，本研究の応用として位置付けられる．これに関しては本年度は国際会議（SC19）での査読付きポスター発表を行った． -
Reduced-precision formats for high-performance and energy-efficient computations
Grant number：16K16062 2016.4 - 2019.3
Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research Grant-in-Aid for Young Scientists (B) Grant-in-Aid for Young Scientists (B)
Mukunoki Daichi
This study explored the possibility of reduced-precision formats which have shorter bit length against the IEEE 32/64 bit floating-point for enchance the performance of numerical computations in terms of both computation speed and energy efficiency. We proposed a light-weight implementation of reduced-precision formats on software and demonstrated the performance improvement, in terms of both speed and energy efficiency, on some data-intensive operations on basic linear algebra.
-
ＧＰＵスパコンのための３倍・４倍精度線形演算ライブラリの開発に関する研究
Grant number：13J01290 2013.4 - 2015.3
日本学術振興会 科学研究費助成事業 特別研究員奨励費 特別研究員奨励費
椋木 大地
本研究の目的は，GPUスパコンにおける3倍・4倍精度演算の実用化を目的として，GPUにおける高性能な3倍・4倍精度線形計算ライブラリの実現に向けた基礎研究を行うことであった．本年度は主として，GPUにおける複数の演算精度に対応した線形計算ライブラリの効率的な実装手法に関する研究を行った，その結果として，複数のNVIDIA GPUアーキテクチャに対応した高速な行列ベクトル積ルーチン（GEMV）の実装手法を開発した．本実装ではGPUにおけるプログラムの実行メカニズムをモデル化し，実行効率が最大となるようなスレッドブロックサイズを自動的に決定するオンライン自動チューニングを採用する．これにより既存の実装と比べ，実行環境や問題サイズに依存して生じる性能の変動を防ぎ，常に高い性能を維持できる．本手法は，ある線形計算を行うプログラム（例えばBLASルーチンなど）において演算精度が異なる複数バージョンを実装・最適化する上で有効であると考えられる．またこの他に，4倍精度演算手法の応用として，倍精度演算性能が単精度演算性能の1/32であるNVIDIA社の最新GPUにおいて，ソフトウェアエミュレーションによる疑似倍精度演算を実装し，倍精度行列積ルーチン（DGEMM）においてハードウェア処理による実装を上回る性能が得られることを示した．本年度に開発したGPU向けソフトウェアの一部は，オープンソースのライブラリとしてウェブ上で公開しており，今後も開発を継続する予定である．
-
Research on high-performance and high-dimensional numerical linear algebra applying an asynchronous task mechanism on the exascale computing era
Grant number：19H04127 2019.4 - 2022.3
Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research Grant-in-Aid for Scientific Research (B) Grant-in-Aid for Scientific Research (B)
Authorship：Coinvestigator(s)
-
Theory and Application of Scalable Numerical Software on an O(100M) core environment
Grant number：15H02709 2015.4 - 2018.3
Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research Grant-in-Aid for Scientific Research (B) Grant-in-Aid for Scientific Research (B)
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.
Teaching Experience (Off-campus) 1
-
情報処理技法（リテラシ）II
2018.9 - 2019.1 （Tokyo Woman's Christian University）