Strong Scalability Analysis of the Albany Land Ice Code on HPC Architectures

High resolution simulations of polar ice sheets play a crucial role in the ongoing effort to develop more accurate and reliable Earth-system models for probabilistic sea-level projections. As part of the DOE Energy Exascale Earth System Model (E3SM), the Albany Land Ice (ALI) code discretizes and solves the First-Order Stokes equations for the ice-sheet velocities. The ALI code has been made portable across different high performance computing (HPC) architectures using Trilinos libraries and the Kokkos programming model but configurations that maximize simulation performance while minimizing losses in architectural efficiency are still under investigation. This study aims to identify an optimal configuration on NERSC Perlmutter and OLCF Frontier through a strong scalability analysis. We show that an MPI+OpenMP configuration performs better on the Perlmutter CPU partition when compared to MPI-only, highlighting the need to improve threaded CPU solvers. We also identify the number of GPUs we can use in our simulations on Perlmutter and Frontier before suffering from significant losses in GPU efficacy.