Background

Large-scale numerical simulations require parallel computers, whose performance for a numerical code is influenced by many factors. It is the purpose of the thesis work to find out how well a particular selection of parallel computers suit for numerical simulations.

What You Will Do

• Compile a large-scale numerical simulation code on different parallel computers
• Test different compilation options
• Do systematic measurement and analysis

What You Will Learn

• Performance tuning/optimization of C/C++ codes
• Use of parallel computers (Origin 3800, IBM Regatta, Linux clusters)

Qualifications

Requirement:

Sufficient programming skills in C and/or C++

Possible advantage:

INF5620