AMEEDA: A general-purpose mapping tool for parallel applications on dedicated clusters

Abstract

The mapping of parallel applications constitutes a difficult problem for which very few practical tools are available. AMEEDA has been developed in order to overcome the lack of a general-purpose mapping tool. The automatic services provided in AMEEDA include instrumentation facilities, parameter extraction modules and mapping strategies. With all these services, and a novel graph formalism called TTIG, users can apply different mapping strategies to the corresponding application through an easy-to-use GUI, and run the application on a PVM cluster using the desired mapping.

This work was supported by the MCyT under contract 2001-2592 and partially sponsored by the Generalitat de Catalunya (G. de Rec. Consolidat 2001SGR-00218).