Repository of exascale problem challenged from the ExCALIBUR programme.
As part of the ongoing ExCALIBUR benchmark activity, we are capturing technical information about the planned use of exascale computing by different research areas. One way in which we are doing this is to work with the ExCALIBUR projects/activities/working groups to develop a set of ExCALIBUR Exascale Challenge Problems. These challenge problems provide a summary of test cases that can be used on an exascale system that demonstrate it is meeting the requirements for the associated research use case. In some, but not all, cases these Challenge Problems can also be baselined on current large HPC systems.
The ExCALIBUR approach draws heavily on work done by the US Exascale Computing Project (ECP).
-
Category 1: Codebase reasonably well-developed for exascale use cases: For these challenge problems, a key concept is the performance baseline, which is a quantitative measure of an application Figure of Merit (FOM) using the fastest computers available at the inception of ExCALIBUR against which the final FOM improvement is measured. In the UK, this includes systems such as ARCHER2 (~25 PFLOP/s) but other systems in other countries of equivalent or larger scale may also be used to define the FOM. For an exascale system we would expect an uplift in the FOM of approximately 40x from a baseline on a 25 PFLOP/s system.
-
Category 2: Codebase needs substantial work for exascale use cases: Without a well-defined starting point at the 10–20 PFLOP/s scale, it is unclear what Figure of Merit (FOM) improvement would correspond to a successful outcome. A more appropriate measure of success for these applications is whether the necessary capability to execute their exascale challenge problems is in place at the end of the project, not the relative performance improvement throughout the project.
- Lattice QCD with Domain-Wall Fermions
- High Performance Geometric Multigrid (HPGMG)
- Simulating the Universe
- To be added