Guppie: A coordination framework for parallel processing using shared memory featuring a master-worker relationship

Sean Christopher McCarthy


Most programs can be parallelized to some extent. The processing power available in computers today makes parallel computing more desirable and attainable than ever before. Many machines today have multiple processors or multiple processing cores making parallel computing more available locally, as well as over a network. In order for parallel applications to be written, they require a computing language, such as C++, and a coordination language (or library), such as Linda. This research involves the creation and implementation of a coordination framework, Guppie, which is easy to use, similar to Linda, but provides more efficiency when dealing with large amounts of messages and data. Greater efficiency can be achieved in coarse-grained parallel computing through the use of shared memory managed through a master-worker relationship.