Clusters are, by now, one of most popular architectures. Hiding latencies as well as getting an optimal assignment of processors, are two issues for many scientific applications, specially if non dedicated clusters are used. Traditionally, high performance scientific applications are parallelized using MPI libraries. Typical implementations of MPI, minimize dynamic features required to face latencies or shared resource usage. Switching from the MPI process model to a threaded programming model in the parallel environment, can help to achieve efficient overlapping and provide abilities for load balancing. Gains achieved by BICAV (our research group's tomographic reconstruction software) in a multithreaded framework, are exposed.
IOS Press, Inc.
6751 Tepper Drive
Clifton, VA 20124
Tel.: +1 703 830 6300
Fax: +1 703 830 2300 firstname.lastname@example.org
(Corporate matters and books only) IOS Press c/o Accucoms US, Inc.
For North America Sales and Customer Service
West Point Commons
Lansdale PA 19446
Tel.: +1 866 855 8967
Fax: +1 215 660 5042 email@example.com