An implementation of tensor product patch smoothers on GPU
arXiv:2405.19004 · doi:10.1137/24M1642706
Abstract
We present a GPU implementation of vertex-patch smoothers for higher order finite element methods in two and three dimensions. Analysis shows that they are not memory bound with respect to GPU DRAM, but with respect to on-chip scratchpad memory. Multigrid operations are optimized through localization and reorganized local operations in on-chip memory, achieving minimal global data transfer and a conflict free memory access pattern. Performance tests demonstrate that the optimized kernel is at least 2 times faster than the straightforward implementation for the Poisson problem, across various polynomial degrees in 2D and 3D, achieving up to 36% of the peak performance in both single and double precision on Nvidia A100 GPU.
References in corpus (9)
- MFEM: a modular finite element methods library
- Nodal Discontinuous Galerkin Methods on Graphics Processors
- Fast matrix-free evaluation of discontinuous Galerkin finite element operators
- Dissecting Tensor Cores via Microbenchmarks: Latency, Throughput and Numeric Behaviors
- GPU performance analysis of a nodal discontinuous Galerkin method for acoustic and elastic models
- Matrix-free multigrid block-preconditioners for higher order Discontinuous Galerkin discretisations
- Multigrid methods for Hdiv-conforming discontinuous Galerkin methods for the Stokes equations
- Efficient low-order refined preconditioners for high-order matrix-free continuous and discontinuous Galerkin methods
- Fast Tensor Product Schwarz Smoothers for High-Order Discontinuous Galerkin Methods