2:30 PM - 2:45 PM
[MGI29-10] Post-Fugaku Feasibility Study and Computational Astrophysics and Planetary Science
Keywords:High-Performance Computing, large-scale simulations
Supercomputer Fugaku was completed in FY2020, and has already been used for three years. While it has achieved excellent results in the area of astrophysics and planetary science, problems have become apparent. MEXT has started a two-year "Feasibility Study on Next-Generation Supercomputing Infrastructures" in FY2022, and the period has been extended to three years. The goal of this study is to clarify the needs of science, industry, and society on the performance and functions, and to propose options for a system that can realize these needs.
Two research teams were selected to study and propose actual systems: one team with RIKEN and the other team with Kobe University. The RIKEN team has been working with multiple vendors to examine various possibilities. The Kobe University team has been studying the system configuration with accelerators, based on the accelerator for deep learning applications that is being developed jointly with PFN.
In this presentation, I present the likely structure of Post-Fugaku system based on open information and discuss the future direction of high-performance systems and the R&D directions for the researchers in the field of astrophysics and planetary science.
Two research teams were selected to study and propose actual systems: one team with RIKEN and the other team with Kobe University. The RIKEN team has been working with multiple vendors to examine various possibilities. The Kobe University team has been studying the system configuration with accelerators, based on the accelerator for deep learning applications that is being developed jointly with PFN.
In this presentation, I present the likely structure of Post-Fugaku system based on open information and discuss the future direction of high-performance systems and the R&D directions for the researchers in the field of astrophysics and planetary science.