pith. sign in

arxiv: 2508.04356 · v3 · submitted 2025-08-06 · 🧮 math.NA · cs.NA

Monolithic Multi-level Overlapping Schwarz Solvers for Fluid Problems

Stephan K\"ohler , Oliver Rheinbach This is my paper

Pith reviewed 2026-05-19 00:42 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords overlapping Schwarzmonolithic preconditionerincompressible flowmulti-level domain decompositionsaddle-point systemsparallel scalabilityGDSW coarse spaceextrusion die geometry
0
0 comments X

The pith

A three-level monolithic overlapping Schwarz preconditioner solves incompressible fluid problems scalably up to 32768 processors on complex geometries.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that extending a monolithic overlapping Schwarz method to three levels, using the generalized Dryja-Smith-Widlund coarse space, delivers good parallel performance for saddle-point systems from incompressible flow. Results cover both a simple Poiseuille flow on the unit cube and a complex extrusion die, with computations reaching 32768 MPI ranks through coupling of the FROSch and FEATFLOW libraries. A sympathetic reader would care because large-scale fluid simulations often require solvers that scale without heavy problem-specific adjustments. The work focuses on numerical and parallel scalability for these additive domain-decomposition methods.

Core claim

We present parallel results up to 32768 MPI ranks for the solution of incompressible fluid problems for a Poiseuille flow example on the unit cube and a complex extrusion die geometry using a two- and a three-level monolithic overlapping Schwarz preconditioner.

What carries the argument

The three-level monolithic overlapping Schwarz preconditioner built on the generalized Dryja-Smith-Widlund coarse space, which treats the velocity-pressure saddle-point structure of the incompressible Navier-Stokes equations in a single monolithic system.

If this is right

  • The method achieves both numerical and parallel scalability for incompressible flow without additional tuning parameters.
  • The same coarse-space construction works for both simple channel flow and complex industrial geometries.
  • Library coupling between overlapping-Schwarz solvers and finite-element flow codes enables these large-scale runs.
  • Extension to three levels preserves the monolithic handling of the velocity-pressure coupling.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The approach may generalize to other saddle-point systems such as those in structural mechanics or magnetohydrodynamics.
  • Further levels could be added to target even larger processor counts on future exascale machines.
  • The observed robustness suggests the coarse space captures essential global modes that simpler two-level methods miss on complex domains.
  • Integration with heterogeneous hardware could reduce wall-clock time by combining the algebraic scalability shown here with accelerator-based local solves.

Load-bearing premise

The generalized Dryja-Smith-Widlund coarse space remains effective when extended from two to three levels for monolithic treatment of incompressible flow saddle-point systems without requiring problem-specific tuning.

What would settle it

An experiment on the extrusion die or a similar geometry where the iteration count or time-to-solution grows sharply when moving from two to three levels or when increasing the processor count beyond a few thousand.

Figures

Figures reproduced from arXiv: 2508.04356 by Oliver Rheinbach, Stephan K\"ohler.

Figure 1
Figure 1. Figure 1: Visualization of the solution of a Poiseuille flow on the unit cube and of the decomposition into 512 subdomains. 4 Numerical Results For our numerical experiments, we consider a Poiseuille flow on the unit cube, to show the parallel performance of our imple￾mentation, and a Poiseuille flow and more advanced geometry, an extrusion die, which can be used in industrial applications. For the Poiseuille flow o… view at source ↗
Figure 2
Figure 2. Figure 2: Visualization of the magnitude of the velocity of a Poiseuille flow on an extrusion die. There are two inflow boundaries and one outflow boundary, see first image from the right: left and bottom boundary are the inflow boundaries and right boundary is the outflow boundary. coarse problem as the coarse problem for the two-level approach for 512 subdomains, which has 5 573 coarse dofs and needs 4.46s for the… view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of the decomposition into 512 subdomains of the extrusion die. 5 Conclusion We implemented successfully a monolithic three-level overlapping Schwarz preconditioner within the Trilinos/FROSch soft￾ware package and tested the scalability with two examples: A Stokes flow problem on the unit cube and on a more complex geometry of an extrusion die. Our numerical experiments show the benefit of the… view at source ↗
read the original abstract

Additive overlapping Schwarz Methods are iterative methods of the domain decomposition type for the solution of partial differential equations. Numerical and parallel scalability of these methods can be achieved by adding coarse levels. A successful coarse space, inspired by iterative substructuring, is the generalized Dryja-Smith-Widlund (GDSW) space. In https://doi.org/10.1137/18M1184047, based on the GDSW approach, two-level monolithic overlapping Schwarz preconditioners for saddle point problems were introduced. We present parallel results up to 32768 MPI ranks for the solution of incompressible fluid problems for a Poiseuille flow example on the unit cube and a complex extrusion die geometry using a two- and a three-level monolithic overlapping Schwarz preconditioner. These results are achieved through the combination of the additive overlapping Schwarz solvers implemented in the Fast and Robust Overlapping Schwarz (FROSch) library https://doi.org/10.1007/978-3-030-56750-7_19, which is part of the Trilinos package ShyLU https://doi.org/10.1109/IPDPS.2012.64, and the FEATFLOW library http://www.featflow.de using a scalable interface for the efficient coupling of the two libraries. This work is part of the project StroemungsRaum - Novel Exascale-Architectures with Heterogeneous Hardware Components for Computational Fluid Dynamics Simulations, funded by the German Bundesministerium fur Forschung, Technologie und Raumfahrt BMFTR (formerly BMBF) as part of the program on New Methods and Technologies for Exascale Computing (SCALEXA).

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The manuscript presents two- and three-level monolithic overlapping Schwarz preconditioners based on the generalized Dryja-Smith-Widlund coarse space for incompressible Navier-Stokes saddle-point systems. It demonstrates parallel scalability up to 32768 MPI ranks on a Poiseuille flow in the unit cube and a complex extrusion-die geometry by coupling the FROSch library (Trilinos/ShyLU) with the FEATFLOW finite-element code through a scalable interface.

Significance. If the reported iteration counts and timing data hold, the work provides concrete evidence that a monolithic three-level GDSW-based overlapping Schwarz method can deliver strong scaling for realistic incompressible flow problems without problem-specific tuning. The explicit algorithmic description of the three-level extension, the library coupling, and the tabulated performance metrics on two geometries constitute a practical contribution to exascale CFD solver development.

minor comments (3)
  1. §4 (numerical results): iteration counts and wall-clock times are presented for both test cases, but the tables would benefit from an additional column reporting the number of degrees of freedom per subdomain at each level to allow direct assessment of the coarse-space size growth.
  2. §3.2 (three-level extension): the description of the inter-level transfer operators is clear, yet a short pseudocode listing the overall preconditioner application would improve readability for readers implementing similar methods.
  3. Figure 5 (extrusion-die geometry): the mesh resolution and partition visualization would be easier to interpret if the number of subdomains and the overlap width were annotated directly on the figure.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive evaluation and recommendation of minor revision. The summary accurately reflects our extension of the two-level GDSW monolithic overlapping Schwarz preconditioners to three levels, the parallel scalability results up to 32768 MPI ranks, and the coupling between FROSch and FEATFLOW for the Poiseuille and extrusion-die test cases. Since the report lists no specific major comments, our response below addresses the overall recommendation.

read point-by-point responses

  1. Referee: No specific major comments provided; recommendation is minor revision.

    Authors: We appreciate the referee's assessment that the work provides concrete evidence for strong scaling of the three-level method on realistic incompressible flow problems. We will perform a minor revision to improve presentation, for example by clarifying the algorithmic description of the three-level extension and ensuring all tabulated performance metrics are fully self-contained. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper's central contribution consists of numerical experiments demonstrating parallel scalability to 32768 MPI ranks for two- and three-level monolithic overlapping Schwarz preconditioners on incompressible flow saddle-point systems. These results are obtained by coupling established libraries (FROSch and FEATFLOW) and extending the generalized Dryja-Smith-Widlund coarse space, with concrete iteration counts and timing data supplied for the Poiseuille and extrusion-die test cases. The reference to prior two-level work supplies algorithmic context but does not bear the load of the new scalability claims, which rest on independent computational verification rather than any self-referential definition, fitted parameter renamed as prediction, or derivation that reduces to the authors' own inputs by construction. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work relies on standard finite-element theory and established domain-decomposition assumptions rather than introducing new fitted parameters or postulated entities.

axioms (1)
  • domain assumption Standard assumptions of overlapping Schwarz theory and inf-sup stable discretizations for incompressible flow hold for the chosen test cases.
    Invoked implicitly when claiming robustness of the GDSW coarse space for saddle-point problems.

pith-pipeline@v0.9.0 · 5836 in / 1275 out tokens · 35653 ms · 2026-05-19T00:42:54.082811+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [1]

    C. R. Dohrmann, A. Klawonn, and O. B. Widlund, A family of energy minimizing coarse spaces for overlapping Schwarz pre- conditioners, in: Domain decomposition methods in science and engineering XVII, , LNCSE, V ol. 60 (Springer, Berlin, 2008), pp. 247–254

    work page 2008

  2. [2]

    C. R. Dohrmann, A. Klawonn, and O. B. Widlund, SIAM J. Numer. Anal. 46(4), 2153–2168 (2008)

    work page 2008

  3. [3]

    Heinlein, A

    A. Heinlein, A. Klawonn, S. Rajamanickam, and O. Rheinbach, FROSch: A fast and robust overlapping Schwarz domain decom- position preconditioner based on Xpetra in Trilinos, in: Domain Decomposition Methods in Science and Engineering XXV , edited by R. Haynes, S. MacLachlan, X. C. Cai, L. Halpern, H. H. Kim, A. Klawonn, and O. Widlund (Springer Internation...

    work page 2020

  4. [4]

    Heinlein, A

    A. Heinlein, A. Klawonn, O. Rheinbach, and F. Röver, A three-level extension of the GDSW overlapping Schwarz preconditioner in two dimensions, in: Advanced Finite Element Methods with Applications: Selected Papers from the 30th Chemnitz Finite Element Symposium 2017, (Springer International Publishing, Cham, 2019), pp. 187–204

    work page 2017

  5. [5]

    Rajamanickam, E

    S. Rajamanickam, E. G. Boman, and M. A. Heroux, Shylu: A hybrid-hybrid solver for multicore platforms, in: 2012 IEEE 26th International Parallel and Distributed Processing Symposium, (May 2012), pp. 631–643

    work page 2012

  6. [6]

    Trilinos public git repository, Web, 2025

    work page 2025

  7. [7]

    Heinlein, C

    A. Heinlein, C. Hochmuth, and A. Klawonn, SIAM Journal on Scientific Computing 41(4), C291–C316 (2019)

    work page 2019

  8. [8]

    Hochmuth, Parallel Overlapping Schwarz Preconditioners for Incompressible Fluid Flow and Fluid-Structure Interaction Prob- lems, PhD thesis, Universität zu Köln, June 2020

    C. Hochmuth, Parallel Overlapping Schwarz Preconditioners for Incompressible Fluid Flow and Fluid-Structure Interaction Prob- lems, PhD thesis, Universität zu Köln, June 2020

    work page 2020

  9. [9]

    Heinlein, A

    A. Heinlein, A. Klawonn, O. Rheinbach, and F. Röver, A three-level extension of the gdsw overlapping schwarz preconditioner in three dimensions, in: International Conference on Domain Decomposition Methods, (2018), pp. 185–192

    work page 2018

  10. [10]

    Röver, Multi-level Extensions for the Fast and Robust Overlapping Schwarz Preconditioners, PhD thesis, Technische Universität Bergakademie Freiberg, 2023

    F. Röver, Multi-level Extensions for the Fast and Robust Overlapping Schwarz Preconditioners, PhD thesis, Technische Universität Bergakademie Freiberg, 2023

    work page 2023

  11. [11]

    Heinlein, O

    A. Heinlein, O. Rheinbach, and F. Röver, Computational Methods in Applied Mathematics 23(4), 953–968 (2023)

    work page 2023

  12. [12]

    FEATFLOW, 2024 (acccessed December 18, 2024)

    work page 2024

  13. [13]

    Heinlein, A

    A. Heinlein, A. Klawonn, and O. Rheinbach, SIAM J. Sci. Comput. 38(6), C713–C747 (2016), Preprint http://tu-freiberg.de/sites/ default/files/media/fakultaet-fuer-mathematik-und-informatik-fakultaet-1-9277/prep/2016-04_fertig.pdf

    work page 2016

  14. [14]

    Heinlein, A

    A. Heinlein, A. Klawonn, and O. Rheinbach, in: Parallel Two-Level Overlapping Schwarz Methods in Fluid-Structure Interaction (Springer International Publishing, 2016), pp. 521–530, Preprint http://tu-freiberg.de/sites/default/files/media/ fakultaet-fuer-mathematik-und-informatik-fakultaet-1-9277/prep/2015-15_fertig_0.pdf

    work page 2016

  15. [15]

    A. Heinlein, Parallel Overlapping Schwarz Preconditioners and Multiscale Discretizations with Applications to Fluid-Structure In- teraction and Highly Heterogeneous Problems, PhD thesis, Universität zu Köln, 2016

    work page 2016

  16. [16]

    Trilinos Project Team, The Trilinos Project Website, 2020 (acccessed May 22, 2020)

    T. Trilinos Project Team, The Trilinos Project Website, 2020 (acccessed May 22, 2020). Copyright line will be provided by the publisher PAMM header will be provided by the publisher 7

    work page 2020

  17. [17]

    Dünnebacke, S

    J. Dünnebacke, S. Turek, C. Lohmann, A. Sokolov, and P. Zajac, The International Journal of High Performance Computing Appli- cations 35(3), 211–225 (2021)

    work page 2021

  18. [18]

    D. Ruda, S. Turek, D. Ribbrock, and P. Zajac, The International Journal of High Performance Computing Applications 36(4), 459– 474 (2022)

    work page 2022

  19. [19]

    Paul, Preconditioning for hyperelasticity-based mesh optimisation, 2017, Preprint

    J. Paul, Preconditioning for hyperelasticity-based mesh optimisation, 2017, Preprint. Copyright line will be provided by the publisher

    work page 2017