The explosive increase in computing power delivered by modern supercomputers promises unprecedented simulations scale and fidelity. Their massively-parallel architectures however pose formidable challenges to algorithm and software development. For example, fully harnessing exascale computers, which will deliver in excess of 1018 operation per second, will require simultaneously executing on the order of a billion operations without being limited by communication and synchronization overhead. This severely constrains the types of simulations that are expected to make efficient use of upcoming exascale architectures, and hence risks limiting their scientific impact in the computational sciences.
The aim of this program is to foster the development of new mathematical tools and formalisms that will enable a new generation of ultra-scalable algorithms for a broad range of applications in computational materials science. Topics of interest will include strategies for scalable single-scale simulations, novel massively-parallel scale-bridging algorithms, and integration of extreme-scale computing into experimental and data science workflows. The program will bring together applied mathematicians, materials scientists, computer scientists, and method developers interested in unlocking the potential of upcoming exascale architectures through novel mathematical approaches.
(University of California, Santa Barbara (UCSB))
Jack Deslippe (Lawrence Berkeley Laboratory)
Virginie Ehrlacher (École Nationale des Ponts-et-Chaussées)
Vikram Gavini (University of Michigan)
Tim Germann (Los Alamos National Laboratory)
Tzanio Kolev (Lawrence Livermore National Laboratory)
Amedeo Perazzo (Stanford University)
Danny Perez (Los Alamos National Laboratory)
Anna Vainchtein (University of Pittsburgh)