Implementation of p-multigrid and approximate fast diagonalization methods in Firedrake

Pablo Brubeck (University of Oxford, 🇬🇧)
Patrick Farrell (University of Oxford, 🇬🇧)
Monday session 2 (Zoom) (15:00–16:30 GMT)

For problems with smooth solutions, high-order methods have very good convergence properties, and in some cases they do not exhibit locking phenomena found in low-order methods. Moreover, due to data-locality and high arithmetic intensity, they are better suited to make efficient use of modern parallel hardware architectures. Unfortunately, the conditioning of the Galerkin matrices is severely affected by \(p\), the polynomial degree of the approximation. In order to obtain practical iterative solvers, we require good preconditioners.

In the \(p\)-variant of multigrid, the problem is often coarsened by rediscretizing on the same mesh with a lower \(p\). We implement a general \(p\)-multigrid (\(p\)-MG) method that can deal with general finite elements and custom coarsening schedules in Firedrake using PETSc. As relaxation, we employ a novel combination of an approximate fast diagonalization method and subspace correction. The scheme is essentially point-block Jacobi in the space of eigenfunctions of a separable approximation to the local stiffness matrix of each cell. The relaxation depends on the tensor-product structure of quadrilateral and hexahedral elements, in a similar manner to sum factorisation.

We employ this relaxation method in two algorithms: a \(p\)-MG preconditioner and a full approximation scheme nonlinear solver. We demonstrate how to combine these two efficiently in a nested iteration with a cascadic outer cycle and inner V-cycles. All available solvers, including geometric and algebraic multigrid, may be employed for the \(p\)-coarse level. The associated computational costs are \(O(p^d)\) to approximate the local stiffness matrix in \(d\) dimensions, and \(O(p^{d+1})\) to apply or update the relaxation, while memory requirements are kept at \(O(p^d)\). We present nonlinear examples such as the \(p\)-Laplacian and incompressible hyperelasticity.