On 27/06/16 13:44, Barry Smith wrote:

Yes, you are correct, there is no MatFDSetType() and in the SetFromOptions we totally cheat http://www.mcs.anl.gov/petsc/petsc-current/src/mat/matfd/fdmatrix.c.html#MatFDColoringSetFromOptions and just stick the value directly into the data structure. For now you can call
PetscOptionsSetValue(NULL,"-mat_fd_type","ds") in your code (before you start using the MatFD stuff) and we'll add the routine to the master branch.

Thanks, that workaround does the trick for now.

I didn't think anyone would care about the difference between ds and wp with MatFD.

It makes a noticeable difference for the problem I'm solving right now. That may be because the problem is not non-dimensionalized (as yet anyway) and so some of the solution elements are of different magnitudes than others. If I non-dimensionalize them it may possibly not matter any more about wp or ds.

- Adrian

-- 
Dr Adrian Croucher
Senior Research Fellow
Department of Engineering Science
University of Auckland, New Zealand
email: a.croucher@auckland.ac.nz
tel: +64 (0)9 923 84611