Joint Seminar: New numerical time stepping algorithms (some of them parallel!) for computer simulations of atmospheric flows

For decades, the rapidly growing computational power of high-performance computing systems has been leveraged to provide increasingly accurate weather forecasts. However, the trend towards massively parallel computing architectures makes translating raw computing power into model performance increasingly difficult. Previously, advances in hardware and computational algorithms contributed to better performance independently. Now, algorithms have to advance in lock-step with hardware to unlock future growth in computational capacity. In my talk, I will present new approaches to time stepping that could improve performance of weather and climate models. I will highlight recent advances in the field of parallel-in-time integration methods that could provide a way to improved parallel performance of models of atmospheric flows but also discuss some of the challenges.

Date

29.10.2019

Time

14:15 h

Place

Geomatikum H3
Lecture Hall H3 (ground floor), University of Hamburg, Bundesstr. 55, Hamburg

Speaker

Daniel Ruprecht, Technische Universität Hamburg

Chair

Bjorn Stevens

Back to listing