Erik Lindahl, Stockholm University
June 28, 2019
Heterogeneous Acceleration and Challenges for Scientific Computing on The Exascale
Modern computer hardware has become tremendously powerful in terms of FLOPS, memory bandwidth, multithreading and accelerators - but while codes could get close to the theoretical peak performance 20 years ago, many of today’s applications struggle to reach 10% due to the complexity of hardware. Here, I will showcase the challenges faced by real-world scientific applications that primarily focus on improving time-to-solution on increasingly powerful supercomputers rather than FLOP-counts, scaling, or relative acceleration. I will discuss how the end of Dennard scaling is a brick wall for many traditional algorithms (which scientists are only beginning to realise), and how we increasingly have been forced to go back and redesign fundamental algorithms that dominated fields as molecular simulation for more than 50 years. Still, with reasonable amounts of effort the future looks exceptionally promising: It is possible to redesign algorithms so they do not only work, but provide outstanding performance on new generations of processors and accelerators. However, this in turn leads to new challenges. To be able to utilise next-generation hardware efficiently many applications will need to move to heterogeneous acceleration where the various compute, memory, storage and network units work on different parts of a problem in parallel, with synchronisation requirements that can sometimes be on the microsecond scale. Finally, I will discuss the strategies needed for all these applications to be able to turn Exascale computing investments into scientific discoveries and industrial impact.
Erik Lindahl received a PhD from the KTH Royal Institute of Technology in 2001, and performed postdoctoral research at Groningen University, Stanford University and the Pasteur Institute. He is currently professor of Biophysics at Stockholm University, with a second appointment as professor of Theoretical Biophysics at the Royal Institute of Technology. Lindahl’s research is focused on understanding the molecular mechanisms of membrane proteins, in particular ion channels, through a combination of molecular simulations and experimental work involving cryo-EM and electrophysiology. He hasauthored some 130 scientific publications and is the recipient of an ERC starting grant.
Lindahl heads the international GROMACS molecular simulation project, which is one of the leading scientific codes to exploit parallelism on all levels from accelerators and assembly code to supercomputers and distributed computing. He is co-director of the Swedish e-Science Research Center as well as the Swedish National Bioinformatics Infrastructure, and lead scientist of the BioExcel Center-of-Excellence for Computational Biomolecular Research. His research work has been awarded with the Prix Jeune Chercheur Blaise Pascal, the Sven and Ebba-Christian Högberg prize, and the Wallenberg Consortium North prize. Lindahl is currently the chair of the PRACE Scientific Steering Committee.
Professor of Biophysics
Dept of Biochemistry & Biophysics