| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 524968 | Parallel Computing | 2006 | 12 Pages |
Abstract
Domain decomposition methods are numerically efficient for the simulation of large scale phenomena; in addition, these algorithms are naturally parallel and feature a good localization of data. Skeleton programming is a high level method for specifying the parallel structure of a program. We describe a full-fledged application of domain decomposition to a three-dimensional flow simulation problem and we express its parallel coordination code with skeleton programming in the purely functional setting of OCamlP3l. We detail the OCamlP3l skeletons (including the new parfun and pardo), their typing, informal semantics, and definition as stream processors. To summarize, functional programming and skeleton programming appear to be an efficient and convenient framework to develop such parallel numerical applications.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Science Applications
Authors
F. Clément, V. Martin, A. Vodicka, R. Di Cosmo, P. Weis,
