TAU-LOP

τ-Lop is a new parallel performance model aimed to help in the design and optimization of parallel algorithms in multi-core clusters. It represents a parallel algorithm and predicts accurately its costs through the concept of concurrent transfers.
By now, τ-Lop has been applied to underlying algorithms in MPICH and Open MPI mainstream implementations of some MPI collectives in shared memory. Current work is in the application of the model to collective operations when deployed in networks of multicore nodes.
Link del activo: http://gim.unex.es/index.php?option=com_content&view=article&id=546&lang=es#Menu