Tuning Spectral Element Preconditioners for Parallel Scalability on GPUs

This paper is concerning runtime tuning spectral element preconditioners for parallel scalability on GPUs. A nascent tuning strategy is proposed to help alleviate the difficulty of choosing a reasonable preconditioner for a given problem. A variety of novel preconditioning schemes are also proposed, including $p$-multigrid with Chebyshev-accelerated Schwarz smoothers, low-order operator preconditioning, and even hybridizing the two approach by treating the low-order operator as the coarse grid operator.

Link to paper

bibtex entry:

@inbook{tuning-sem-preco,
author = {Malachi Phillips and Stefan Kerkemeier and Paul Fischer},
title = {Tuning Spectral Element Preconditioners for Parallel Scalability on GPUs},
booktitle = {Proceedings of the 2022 SIAM Conference on Parallel Processing for Scientific Computing (PP)},
chapter = {},
pages = {37-48},
doi = {10.1137/1.9781611977141.4},
URL = {https://epubs.siam.org/doi/abs/10.1137/1.9781611977141.4},
eprint = {https://epubs.siam.org/doi/pdf/10.1137/1.9781611977141.4},
}