Implementing Hardware-Agnostic Large-Scale Tiled Linear Algebra: Lessons in High Performance Computing Accessibility
Decanato - Facoltà di scienze informatiche
Data: 5 Luglio 2024 / 10:30 - 11:30
USI East Campus, Room D0.02
Speaker: Evelyne Ringoot, Massachusetts Institute of Technology
Abstract: Computer science has embedded itself in almost every scientific field: simulations are becoming a requirement for almost any theoretical or experimental study, and dataset sizes have exponentially grown. At the same time, computational tools have not necessarily become easier to use, as the field of High-Performance Computing (HPC) is continuously advancing. This work aims to increase HPC accessibility for all scientists. It proposes a hardware-agnostic multi-scale Tiled Linear Algebra abstraction that enables scientists to take advantage of a less tedious software development process for testing new algorithms. We demonstrate the utility of the abstraction by implementing the tiled Singular Value Decomposition.
Biography: Evelyne Ringoot is a PhD student in the departments of Computational Science and Engineering and Mathematics at the Massachusetts Institute of Technology. She works in the Julialab with Professor Alan Edelman on High-Performance Computing (HPC) implementations for large-scale numerical linear algebra. She previously obtained a MSc in Civil Engineering at the Vrije Universiteit Brussel/Université Libre de Bruxelles, was a visiting student at EPFL Lausanne and Ulsan UNIST, and has industry experience in data analytics consulting. Her research aims to make HPC accessible across scientific fields.
Host: Prof. Olaf Schenk
