Abstract
Clusters of SMPs (symmetric multiprocessors) gain more and more importance in high performance computing and are often used with a hybrid programming model based on message passing and multithreading. For irregular application programs with dynamically changing computation and data access behavior a flexible programming model is needed to achieve load balance and efficiency. We have proposed task pool teams as a new hybrid concept to realize irregular algorithms on clusters of SMPs efficiently. Task pool teams offer implicit load balance on single cluster nodes together with multithreaded, MPI-based communication for the exchange of remote data. This paper introduces the programming model and library support with an easy to use application programmer interface for task pool teams.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have