LFU:online

Students who have completed this course will be able to describe and classify relevant computer architectures and programming models in HPC. They will be able to identify, employ and adapt relevant numerical methods for the solution of linear PDEs in general and the diffusion equation in particular. Students will be able to provide parallel implementations for these methods and evaluate them with regard to their performance and maintainability.

Introduction to the most important and basic concepts of parallel computing, types and properties of parallel computers, performance metrics such as speed-up and efficiency, programming in OpenMP or MPI, work and data distribution, program optimisation. These topics will be illustrated and practically applied using the example of the Gauss-Seidel scheme applied to the diffusion equation.

Theory sessions coupled with practical exercises

The course will focus on the development of a parallel solver for the diffusion equation to be written in C++.

The grade will be determined based on ongoing participation and continuous assessment of the contributions of the students. This will include:

- questions and contributions to discussions during the course

- exercises

- progress in the implementation of the different methods of the main course project

Literature will be provided during the course.

This course assumes familiarity with general programming in C++ and a scripting language preferred by the students. Also knowledge about basic software version control will be very helpful.