Applications of the OpenCAL scientific library in the context of CFD: Applications to debris flows

摘要

We here present computational results as performed by the new OpenCAL scientific library for seamless implementation of models referred to Computational Fluid Dynamics (CFD). In particular, OpenCAL allows for the straightforward definition of Cellular Automata- and Finite Differences-based simulation models of complex systems, by also supporting Extended Cellular Automata and all those computing paradigms based on regular computational grids. Two different parallel implementations are presented, which permit to exploit both multi-core CPUs on shared memory computers, and a wide range of heterogeneous devices like GPUs and other many-core coprocessors. Experiments, referred to a well-known cellular automata, namely the SCIDDICA S model for landslide simulation, exhibit good scalability, and numerical correctness and efficiency have been assessed on both multi-core CPUs and different many-core coprocessors. The good performances that were achieved in all benchmarks confirm the library's suitability for model development of complex systems and their execution on parallel heterogeneous computational devices.