This is a read only copy of the old FEniCS QA forum. Please visit the new QA forum to ask questions

PETSc fieldsplit preconditioner for Stokes

+3 votes

PETSc offers a number of good preconditioners for the Stokes problem, see http://www.mcs.anl.gov/petsc/documentation/tutorials/ACTSTutorial2012.pdf (p. 81 ff.), which all seem to be accessible with only PETSc options.

Ho is it possible to leverage those preconditioners?
I release that there is PETScOptions.set('pc_type', 'fieldsplit') and more, but whenever I set this particular options I get

*** Reason:  PETSc error code is: 77.
*** Where:   This error was encountered inside /home/nschloe/Work/FEniCS/nightly/src/dolfin/dolfin/la/PETScKrylovSolver.cpp

Does anyone have this running?

asked Oct 30, 2013 by nschloe FEniCS User (7,120 points)

p.79 says that you need to use ISes to define fields. Have you done this? And is there a facility for this in DOLFIN?

commented Nov 1, 2013 by Jan Blechta FEniCS Expert (51,420 points)

I checked out Garth's branch and see what I can do.

commented Nov 1, 2013 by nschloe FEniCS User (7,120 points)

1 Answer

+3 votes
 
Best answer

See example programs at https://bitbucket.org/magma-dynamics/preconditioning/. The interface in DOLFIN will improved as I get a clearer idea of how the interface should look. Some optimisation in DOLFIN are also underway.

answered Dec 7, 2013 by Garth N. Wells FEniCS Expert (35,930 points)
edited Dec 10, 2013 by Garth N. Wells

Hi, has the Dolfin interface been improved as you mention?

commented Sep 4, 2014 by mwelland FEniCS User (8,410 points)
...