Numerical Computation Algorithms for Large-scale Parallel Environment
Project Area | Materials Design through Computics: Complex Correlation and Non-equilibrium Dynamics |
Project/Area Number |
22104003
|
Research Category |
Grant-in-Aid for Scientific Research on Innovative Areas (Research in a proposed research area)
|
Allocation Type | Single-year Grants |
Review Section |
Science and Engineering
|
Research Institution | University of Tsukuba |
Principal Investigator |
|
Co-Investigator(Kenkyū-buntansha) |
IMAMURA Toshiyuki 理化学研究所, 計算科学研究機構, チームリーダー (60361838)
TADANO Hiroto 筑波大学, システム情報系, 助教 (50507845)
|
Co-Investigator(Renkei-kenkyūsha) |
SATO Mitsuhisa 筑波大学, システム情報系, 教授 (60333481)
BOKU Taisuke 筑波大学, システム情報系, 教授 (90209346)
SAKURAI Tetsuya 筑波大学, システム情報系, 教授 (60187086)
|
Project Period (FY) |
2010-04-01 – 2015-03-31
|
Project Status |
Completed (Fiscal Year 2014)
|
Budget Amount *help |
¥22,490,000 (Direct Cost: ¥17,300,000、Indirect Cost: ¥5,190,000)
Fiscal Year 2014: ¥3,510,000 (Direct Cost: ¥2,700,000、Indirect Cost: ¥810,000)
Fiscal Year 2013: ¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2012: ¥3,900,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥900,000)
Fiscal Year 2011: ¥3,900,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥900,000)
Fiscal Year 2010: ¥7,410,000 (Direct Cost: ¥5,700,000、Indirect Cost: ¥1,710,000)
|
Keywords | 大規模並列環境 / 数値計算アルゴリズム / GPGPU / 高速フーリエ変換 / 3倍精度浮動小数点演算 / 固有値ソルバ / Block Krylovアルゴリズム / 4倍・8倍精度演算 / 疎行列ベクトル積 / Block Krylov部分空間反復法 / Sakurai-Sugiura法 |
Outline of Final Research Achievements |
Research was conducted on the use of large-scale parallel environments, whose performance exceeded 10 petaflops, for numerical computation algorithms. These included the fast Fourier transform (FFT), quadruple- and octuple-precision basic linear algebra subprograms (BLAS) on graphics processing units (GPUs), triple- and quadruple-precision floating point operations on GPUs, sparse matrix-vector multiplication on GPUs, and a real symmetric eigenvalue solver operating in a CPU + GPU environment. An evaluation was also carried out of the performance of a newly developed eigenvalue solver and an existing solver in a GPU environment, a high-performance eigenvalue solver employing third-generation NVIDIA GPUs, solution acceleration of simultaneous linear algebraic equations using the Block Krylov algorithm, accuracy improvement of Block Krylov partial space iterative methods, and performance-tuning methods.
|
Report
(6 results)
Research Products
(32 results)