High-Productivity Computing in Computational Physics Education


  Guy Tel-Zur  
BGU
NRCN

I describe the development of a new course in Computational Physics at the Ben-Gurion University.

This elective course for 3rd year undergraduates and MSc. students is being taught during one semester.

Computational Physics is by now well accepted as the Third Pillar of Science. This paper’s claim is that modern Computational Physics education should deal also with High-Productivity Computing.

The traditional approach of teaching Computational Physics emphasizes “Correctness” and then “Accuracy” and I add also "Performance".

Along with topics in Mathematical Methods and case studies in Physics the course deals a significant amount of time with “Mini-Courses” in topics such as: High-Throughput Computing - Condor, Parallel Programming - MPI and OpenMP, How to build a Beowulf, Visualization and Grid and Cloud Computing.

The course does not intend to teach neither new physics nor new mathematics but it is focused on an integrated approach for solving problems starting from the physics problem, the corresponding mathematical solution, the numerical scheme, writing an efficient computer code and finally analysis and visualization