Pore-Network Flow Simulator on Generic Hybrid-Parallel Computational Platforms
Hybrid-parallelism provides tremendous speed up compared to singular distributed or shared memory parallelism. A porous flow model constitutes data crunching of enormous amounts which calls for advanced parallel programming models for successful simulations. Such has been developed by our multi-scale multi-physics group at Univeristy of Texas at El Paso with the help of Sandia National Labs's Trilinos and Kokkos library on a generic programming platform. The flow simulator is now under testing on massively parallel architectures with a hybrid distributed and shared memory model.
