Moderator: Rosa M. Badia (Barcelona Supercomputing Center (BSC), Polytechnic University of Catalonia)
Panelists: Rajeev Thakur (Argonne National Laboratory (ANL)), Samuel Thibault (University of Bordeaux (IMB); National Institute for Research in Computer Science and Automation (Inria), France), Didem Unat (Koç University, Turkey), Mitsuhisa Sato (RIKEN Center for Computational Science (R-CCS)), Vicenç Beltran (Barcelona Supercomputing Center (BSC))
Abstract: With exascale computers around the corner, the next challenge to consider is zettascale systems. At this moment, it is difficult to predict how a zettascale machine will be built. It has been forecast to be composed of more diverse and heterogeneous components. Different types of accelerators are expected, including new paradigms like quantum computing. These new systems will pose major scalability issues and severely limit the types of parallelism that can be exploited.
Additionally, the applications to be executed are also expected to grow in complexity and size in order to leverage the existing computing infrastructure and enable progress in multiple areas of science, technology and engineering. While multiple components of the software stack will be relevant for these systems to perform efficiently, the runtime is initially identified as a critical one. With such a scenario in mind, this panel aims to debate how future runtime systems will look.