Parallel Scientific Computing with Applications in Porous Media:
Dr. Andreas Yiotis

This course is an introduction to the basic concepts and techniques of parallel scientific computing using the standard Message Passing Interface (MPI) and OpenMP protocols with applications on massively parallel supercomputer systems. These protocols are widely used for the computationally efficient solution of large-scale computational scientific and engineering problems, including those of aerodynamics around vehicles, planes and complex structures (i.e. wind turbines), the dynamics of buildings and bridges, in weather forecasting and in complex porous media. The course includes an introduction to the characteristics and differences between shared- and distributed-memory computational systems, with particular emphasis on the basic components of distributed-memory systems and their individual role. The main structure of a parallel code is analyzed with respect to the standard serial approach, and typical domain decomposition strategies are presented. The main point-to-point and collective communication routines among computational nodes are presented, and their specific usage in selected applications is analyzed. The course includes also hands-on practice in the development of numerical modeling algorithms for the efficient solution of large-scale systems of linear equations for scientific computing applications.