MFEM: accelerating efficient solution of PDEs at exascale
Tzanio Kolev (Lawrence Livermore National Lab)
https://mfem.org
kolev1@llnl.gov
tzanio
Exascale architectures require rethinking of the numerical algorithms used in large-scale PDE-based applications. These architectures favor algorithms, such as high-order finite elements, that expose fine-grain parallelism and maximize the ratio of floating point operations to energy intensive data movement. In this talk we present an overview of MFEM, a scalable library for high-order finite element discretization of PDEs on general unstructured grids. We also report on the work in the Center for Efficient Exascale Discretizations, a co-design center in the US Exascale Computing Project focused on next-generation discretization software and algorithms. We will describe recent research on performance optimizations for GPU architectures, scalable meshing and discretization algorithms, and matrix-free preconditioning.