Kursusinformation

The course is fully booked! Please sign up at the waiting list. 

For inquiries regarding registration, cancellation or waiting list, please contact the PhD administration, Katrine Søndergaard kes@adm.aau.dk.

This is a course in scientific computing using the popular programming language Python. Python is popular in science due to a number of advantages such as having a rich set of libraries for computing and data visualization, excellent performance-optimizing possibilities, standard tools for simple parallel computing, fast development cycle and high productivity – just to name a few. Python is open source and as such an asset for any researcher following the reproducible research paradigm.

This part of the course covers the main area: High performance computing.

High performance computing:

1. High-performance computing and computer architectures
2. Performance optimization
   1. Numba (just in time compilation)
   2. Numba GPU functionality
   3. Other options - briefly: Cython (compiled Python via C-extensions), f2py (inclusion of Fortran code in Python)
3. Parallel/distributed computing
   1. Theoretical aspects (Amdahl's and Gustafson-Barsis' law)
   2. Parallel computing on one computer
   3. Distributed computing across multiple computers

Audience: The targeted audience is all PhD students with some experience in Python programming and an interest in developing high-performance Python code for various scientific computing purposes. By this we mean code to solve actual problems where [lots of] floating-point computations are needed.

Prerequisites: Participants must have some experience in programming Python. If not, there is an introductory course "Scientific Computing using Python - 1. Python + Scientific Computing". Further, some basic skills in general use of a computer are expected. The tools applied work best using Linux or Mac OSX – Microsoft Windows may experience challenges when using parallel computing.

Criteria for assessment: A standard mini-project must be delivered (4-8 pages description) in addition to the developed code. The code must include testing/validation, and performance evaluation of parallel computing. An acceptable mini-project and at-least 75% participation is required to pass the course.

Learning objectives: After completing the course the participants will:

• Know how to use methods and software for performance optimization.
• Know when and how to apply parallel computing for scientific computing.