Abstract
- Achieving performance portability across numerous current and future architectures is a challenge. The Kokkos programming model allows for a single implementation that can achieve good on-node parallel performance for diverse architectures by capitalizing on the flexibility of data structure allocation and access.
- Climate projections need to be informed by ensembles of runs to establish the probabilistic. While launching ensembles as an outer loop is trivially-parallel, superior speed-ups can be gained by embedding the ensembles as an inner loop instead.
- Implicit methods have promise in some climate components to achieve better scalability for high resolution discretizations. Analytic Jacobian matrices can be calculated efficiently with little code development time.
All three of these potentially-transformative capabilities (1) are enabled throughout C++ templates to swap out data types in single common code base and (2) will already exist in ACME v1 in the MPAS-LI Albany/FELIX solver.