dealiiX Seminar: Efficient finite element solvers with the matrix-free infrastructure of deal.II
This dealiiX seminar explores how the matrix-free infrastructure of deal.II enables faster and more scalable finite element solvers, improving performance for large-scale and high-order simulations.
When: 7 April 2026
Where: Online via Zoom
For many finite-element problems, the solution of nonlinear or linear systems takes the largest share of computing time. In order to increase the performance of this crucial algorithmic component, the deal.II library provides a high-performance matrix-free infrastructure. Instead of assembling large sparse matrices that get used within iterative solvers, it directly applies operators in a matrix-free way. The talk explains how this approach improves memory efficiency, cache utilisation, and scalability, especially for high-order discretisations and large-scale parallel computations. It also highlights practical solver and preconditioner strategies within deal.II that enable robust performance for challenging partial differential equation problems on contemporary hardware.
Speaker: Prof. Martin Kronbichler (Ruhr University Bochum)
