Parallel compact algorithms for the subsurface scattering problems

Yury Gryazin, Idaho State University

Usage Details

This project is part of the XSEDE Scholars Program. The main goal of the project is to introduce graduate students to the basics of the efficient parallel implementation of the state of the art numerical algorithms on the cutting edge supercomputer architecture. This course is designed to complement the graduate seminar on the Krylov subspace iterative methods. In this project we will discuss the major steps and challenges which a computational mathematician or physicist faces during the implementation of the developed numerical algorithms on the parallel supercomputers or clusters. The graduate students in the group have some experience with the theory of advance numerical linear algebra, nonlinear convex optimization, finite volume and finite elements methods. They also have some basic knowledge of modern computer architecture and parallel programming. But neither of them have any experience with the practical parallel implementation of the developed numerical technique. The Blue Waters project will provide them with opportunity to gain this experience and to successfully complete their MS research projects. In this course we will discuss such practical issues as: 1. How to configure the necessary software components before starting the programming work on implementation of the developed parallel algorithms; 2. How to write simple make files; 3. How to use available parallel libraries (e.g., Intel MKL libraries); 4. How to profile the developed code; 5. How to conduct the time complexity analysis of the developed parallel algorithms; and 6. How to optimize the parallel code. At the end of the course the graduate students should feel comfortable with the parallel implementation of the numerical methods on parallel computers.