Practical parallelization of scientific applications with OpenMP, OpenACC and MPI

This work aims at distilling a systematic methodology to modernize existing sequential scientific codes with little re-designing effort, turning an old codebase into modern code, i.e., parallel and robust code. We propose semi-automatic parallelize applications designed purely programming mindset, possibly using global variables, aliasing, random number generators, stateful functions. demonstrate that the same works for parallelization in shared memory model (via OpenMP), message passing MPI), General Purpose Computing on GPU OpenACC). The method is demonstrated parallelizing four real-world domain of physics material science. itself has been distilled collaboration MSc students Parallel course University Torino, applied it first time project they presented final exam course. Every year hosts some special lectures from industry representatives, who present how use computing offer be parallelized.