SNESNGS#

Either calls the user-provided Gauss-Seidel solution routine provided with SNESSetNGS() or does a finite difference secant approximation using coloring [BKST15].

Options Database Keys#

-snes_ngs_sweeps - Number of sweeps of nonlinear GS to apply
-snes_ngs_atol - Absolute residual tolerance for nonlinear GS iteration
-snes_ngs_rtol - Relative residual tolerance for nonlinear GS iteration
-snes_ngs_stol - Absolute update tolerance for nonlinear GS iteration
-snes_ngs_max_it - Maximum number of sweeps of nonlinea GS to apply
-snes_ngs_secant - Use pointwise secant local Jacobian approximation with coloring instead of user provided Gauss-Seidel routine, this is used by default if no user provided Gauss-Seidel routine is available. Requires either that a DM that can compute a coloring is available or a Jacobian sparse matrix is provided (from which to get the coloring).
-snes_ngs_secant_h - Differencing parameter for secant approximation
-snes_ngs_secant_mat_coloring - Use the graph coloring of the Jacobian for the secant GS even if a DM is available.
-snes_norm_schedule <none, always, initialonly, finalonly, initialfinalonly> - how often the residual norms are computed

Notes#

the Gauss-Seidel smoother is inherited through composition. If a solver has been created with SNESGetNPC(), it will have its parent’s Gauss-Seidel routine associated with it.

By default this routine computes the solution norm at each iteration, this can be time consuming, you can turn this off with SNESSetNormSchedule() or -snes_norm_schedule none

References#

[BKST15]

Peter R. Brune, Matthew G. Knepley, Barry F. Smith, and Xuemin Tu. Composing scalable nonlinear algebraic solvers. SIAM Review, 57(4):535–565, 2015. http://www.mcs.anl.gov/papers/P2010-0112.pdf. URL: http://www.mcs.anl.gov/papers/P2010-0112.pdf, doi:10.1137/130936725.

Level#

advanced

Location#

src/snes/impls/gs/snesgs.c

Index of all SNES routines
Table of Contents for all manual pages
Index of all manual pages