Docentship Lecture: Accuracy, Reliability, and Energy Efficiency in Scientific Computing
- Date: 29 November 2023, 10:15–11:15
- Location: Ångström Laboratory, Room 101190
- Type: Docent trial lecture
- Lecturer: Roman Iakymchuk
- Organiser: Department of Information Technology
- Contact person: Roman Iakymchuk
In scientific computing, we solve problems expressed by equations with the help of computers. Often, we employ supercomputers built of hundreds or thousands of processors to solve challenging and larger problems. On computers, in general, we operate with the finite representation of numbers and operations that lack associativity, which is particularly noticeable in parallel environments.
This lecture will commence with the introduction to finite precision and computer arithmetic. Then, I will outline strategies for more accurate and reproducible computations, especially in parallel environments. Finally, I will glimpse through an approach for energy-efficient computing that is motivated by the energy consumption constraint for modern supercomputers.
The lecture is an obligatory teaching test for those applying for admittance as docent and it should be possible for students and others with basic academic education in the relevant field to follow it. The lecture lasts 40-45 minutes with subsequent discussion. The lecture will be given in English.