1: \begin{abstract}

2: \small\baselineskip=9pt

3: The Poisson pressure solve resulting from the spectral element discretization

4: of the incompressible Navier-Stokes equation requires fast, robust, and

5: scalable preconditioning.  In the current work, a parallel scaling study of

6: Chebyshev-accelerated Schwarz and Jacobi preconditioning schemes is presented,

7: with special focus on GPU architectures, such as OLCF's Summit.  Convergence

8: properties of the Chebyshev-accelerated schemes are compared with alternative

9: methods, such as low-order preconditioners combined with algebraic multigrid.

10: Performance and scalability results are presented for a variety of

11: preconditioner and solver settings. The authors demonstrate that

12: Chebyshev-accelerated-Schwarz methods provide a robust and effective smoothing

13: strategy when using $p$-multigrid as a preconditioner in a Krylov-subspace

14: projector.

15: The variety of cases to be addressed, on a

16: wide range of processor counts, suggests that

17: performance can be enhanced by automated run-time selection of the

18: preconditioner and associated parameters.

19: \end{abstract}

20: