In this work efficiency of some parallel realizations of an algorithm for the numerical solution of a boundary-value problem for the shallow water equations which were performed with the help of the MPI library for C language is compared. The first approach is based on the decomposition of a computational domain without overlapping subdomains, the second approach is based on the decomposition with shady sides. Theoretical estimates of acceleration for the parallel algorithm are given.
Numerical results for a model grid and nonstructured grid for the Okhotsk sea are presented. Results concerning acceleration of computations depending on the number of processes, the type of communication realization, and the method of the decomposition of a computational domain are presented.
The work was supported by Russian Foundation of Fundamental Research (grant 08-01-00621-à) and the President Program "leading scientific schools of Russian Federation (grant ÍØ-3431.2008.9)
Note. Abstracts are published in author's edition
Comments |
[Home] [Conference] |
© 1996-2000, Institute of computational Techologies SB RAS, Novosibirsk
© 1996-2000, Siberian Branch of Russian Academy of Science, Novosibirsk