On Wed, Oct 24, 2012 at 6:10 PM, Dmitry Karpeev <karp...@mcs.anl.gov> wrote: > Hi All, > > I just pushed a change tweaking the nonlinear_solver API a bit: it allows > the user to provide callbacks that calculate > the nullspace or the near-nullspace for the system's Jacobian, which can be > used in solving a degenerate problem (e.g., using KSP) > or to help a suitably-equipped AMG preconditioner construct a better coarse > space, respectively. At the moment the only consumer > of this functionality is in Moose, but I'll try to come up with a native > libMesh example soon. This API change follows the pattern > of other similar callbacks (e.g., residual), is fully backward compatible > and #ifdef protected, so it's only compiled with petsc-3.3 or higher. > I've tested it (i.e., ran examples) with petsc-3.3 and petsc-3.2, and all > seems to be well, but let me know if this causes any kind of grief > or if there are any objections to this change.
FYI, I just committed a patch that changes the name of your function from buildMatNullSpace() -> build_mat_null_space() to be more consistent with LibMesh's naming scheme, and does some reformatting. -- John ------------------------------------------------------------------------------ Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_sfd2d_oct _______________________________________________ Libmesh-devel mailing list Libmesh-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-devel
