Matrix-Free Local Multigrid for C¹ Hermite Discretizations on GPUs

DOI10.48550/ARXIV.2610.30308MaRDI QIDQ7401828arXivFDO

Authors Michał Tomasz Wichrowski, Michał Wichrowski

Publication date 21 October 2026

Published in arXiv

Tldr We present a matrix-free local multigrid method for C^1 Hermite discretizations on adaptively refined Cartesian meshes using a cell-wise primal-dual formulation. The method enforces continuity without forming a constraint operator, eliminating the need for hanging nodes or edge operators. On an NVIDIA A100, we obtain efficient solvers with Chebyshev-Jacobi and vertex-patch smoothers, with up to 4.23 million cells solved per second on uniform meshes.

Full work available at URL https://doi.org/10.48550/ARXIV.2610.30308

Mathematics Subject Classification ID

Preconditioners for iterative methods (65F08) Finite element, Rayleigh-Ritz and Galerkin methods for boundary value problems involving PDEs (65N30) Multigrid methods; domain decomposition for boundary value problems involving PDEs (65N55) Complexity and performance of numerical algorithms (65Y20)

arXiv classification

math.NA

Cites work

• \(C^{0}\) interior penalty methods
• A family of C¹ finite elements with optimal approximation properties for various Galerkin methods for 2nd and 4th order problems
• A fast algorithm for spectral differentiation
• A Flexible, Parallel, Adaptive Geometric Multigrid Method for FEM
• A generic interface for parallel cell-based finite element operator application
• A novel implementation of Hermite finite element methods in the deal.II library
• A Scalable and Robust Vertex-Star Relaxation for High-Order FEM
• Adaptive multilevel methods with local smoothing for \(H^1\)- and \(H^{\text{curl}}\)-conforming high order finite element methods
• An implementation of tensor product patch smoothers on GPU
• An implementation of tensor product patch smoothers on GPUs
• An Interior Penalty Finite Element Method with Discontinuous Elements
• An optimal order multigrid method for biharmonic, \(C^ 1\) finite element equations
• Approximate tensor-product preconditioners for very high order discontinuous Galerkin methods
• Coalesced Matrix-Free Finite Elements in Cell-Wise Storage
• Continuous/discontinuous finite element approximations of fourth-order elliptic problems in structural and continuum mechanics with applications to thin beams and plates, and strain gradient elasticity
• Convergence of Multigrid Algorithms for Interior Penalty Methods
• Direct solution of partial difference equations by tensor product methods
• Divergence-conforming discontinuous Galerkin methods and C^0 interior penalty methods
• Eigenvalues and eigenvectors of symmetric centrosymmetric matrices
• Fast matrix-free evaluation of discontinuous Galerkin finite element operators
• Fast tensor product Schwarz smoothers for high-order discontinuous Galerkin methods
• Fully discrete hp-finite elements: Fast quadrature
• Hybrid multigrid/Schwarz algorithms for the spectral element method
• Implementing Spectral Methods for Partial Differential Equations
• Isogeometric Schwarz preconditioners for the biharmonic problem
• Iterative Methods by Space Decomposition and Subspace Correction
• Multigrid Algorithms for C0 Interior Penalty Methods
• Multilevel interior penalty methods on GPUs
• Multilevel Interior Penalty Methods on GPUs
• Spectral methods for problems in complex geometries
• Symmetric centrosymmetric matrix-vector multiplication
• Tensor-product vertex patch smoothers for biharmonic problems
• The Fast Adaptive Composite Grid (FAC) Method for Elliptic Equations
• The Finite Element Method for Elliptic Problems
• The TUBA Family of Plate Elements for the Matrix Displacement Method
• Two-level non-overlapping Schwarz preconditioners for a discontinuous Galerkin approximation of the biharmonic equation
• Unified Analysis of Discontinuous Galerkin Methods for Elliptic Problems

This page was built for publication: Matrix-Free Local Multigrid for $C^1$ Hermite Discretizations on GPUs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7401828)