Speaker: Prof. Dr. Ulrich Rüde, Chair for System Simulation (LSS), FAU

Slides: https://hpc.fau.de/files/2025/02/2025-02-04-Perflab-Ruede.pdf

Abstract

This talk concerns the efficiency of algorithms and their implementation. When developing a new solver, your goal will naturally be to create an efficient method. This is alone because your prospects for publication would be greatly diminished if you were to advertise your new method as inherently inefficient. Efficiency, therefore, is central to research in the field. However, upon closer examination, we find that it is a surprisingly ambiguous and poorly defined concept. This observation prompts us to delve deeper: What, in fact, is efficiency? How do we measure it? How can we determine whether an algorithm is efficient? Is it the algorithm itself that must be efficient, or is it the implementation? Or, perhaps, is it not efficient at all?