Parallelizable Block Rosenbrock Methods for Linear Variable-coefficient System of ODEs(Algorithm & Numerical Computation)
スポンサーリンク
概要
- 論文の詳細を見る
In the previous paper (Esaki and Mitsui, 2001), we proposed parallelizable ROW-type discretization methods to apply to a linear variable-coefficient system of ODEs. They showed good performance on the parallel computing system, but, since the maximum order cannot exceed three, they are considered not to be much practical. In the present paper, we develop a generalized implicit Runge-Kutta method and its block upper-triangular form, named as a block Rosenbrock method, and derive a parallelizable one. Order analysis, global convergence and stability analysis are carried out for the fourth order scheme of the new method. Numerical experiments show its practicality under a parallel computer environment by comparing other conventional methods.
- 一般社団法人情報処理学会の論文
- 2004-10-15
著者
-
Mitsui T
Graduate School Of Information Science Nagoya University
-
Mitsui Taketomo
Graduate School Of Human Informatics Nagoya University
-
ESAKI NOBUYUKI
Toyota National College of Technology
関連論文
- Parallel Predictor-Corrector Iteration of Pseudo Two-Step RK Methods for Nonstiff IVPs
- A Class of Explicit Parallel Two-Step Runge-Kutta Methods
- Collocation-based Two-step Runge-Kutta Methods
- Parallelizable Block Rosenbrock Methods for Linear Variable-coefficient System of ODEs(Algorithm & Numerical Computation)
- Parallelizable ROW Methods for Linear Variable-coefficient System of ODEs (ハイパフォーマンスコンピューティングシステム) -- (数値解析・並列数値処理)
- Unique Solvability of Nonlinear Systems Arising in the Multiderivative Block Methods
- Predator-Prey Dynamics with Delay when Prey Dispersing in n-Patch Environment
- Finite-band Solutions of the Dirac Soliton Equation throrugh a Reduction Technique
- Stability of Linear Delay Differential Systems with Matrices Having Common Eigenvectors