Re: [firedrake] firedrake Digest, Vol 32, Issue 9

Hi Sander,

On 27 Aug 2019, at 20:46, Sander Rhebergen <srhebergen@uwaterloo.ca> wrote:

Hi Matt,

Thanks for the explanation. I'm new to this formulation as well and not sure about the best way to address the solver part, but I will try out your suggestions.

Sander

[...] It is sometimes possible to hit mumps hard enough to handle rank deficient systems. Perhaps something like -pc_type lu -pc_factor_mat_solver_type mumps And then additionally -mat_mumps_icntl_24 1 # enable detection of null pivots -mat_mumps_icntl_25 0 # return a possible solution See Section 5.9 of the mumps user manual for more: http://mumps.enseeiht.fr/doc/userguide_5.2.1.pdf Cheers, Lawrence