S. McIntosh-Smith, University of Bristol (UK)
Abstract: Exascale systems are expected to be exotic creatures, with heterogeneous architectures, millions of degrees of parallelism, and deep memory hierarchies. What are the most promising programming models for these Exascale machines? Is it inevitable that MPI is going to be used for inter-node parallelism while OpenMP, Kokkos or similar models are used within the nodes? Or will other approaches, such as PGAS finally have their day? The talk explored the options, and reviewed what’s already known about the programming models announced for Exascale systems around the world. Some predictions and recommendations for developers of scientific codes aiming for Exascale were also presented.