On Wed, 16 Nov 2016, Salazar De Troya, Miguel wrote:
LinearSolver does not have the capability to have null spaces attached to
it. In my opinion, wouldn’t it be a good idea to have a Solver class from
which all the solvers can inherit common things like attaching null spaces
or setting up the preconditioner?
In this case, specifically? IMHO it would.
My own concern is that that adding some of these APIs at the base
class level would be a lie. Nobody is going to be implementing null
space attachment for Trilinos any time soon, or for Laspack, or for
Eigen, so putting an API in the base class that suggests users can use
null space attachment in any solver is misleading.
However, you may have noticed that the APIs we have already suffer
from this problem, living in NonlinearSolver rather than just
PetscNonlinearSolver; moving them back to a Solver (I'd call it
AlgebraicSolver) class wouldn't make things any worse.
---
Roy
------------------------------------------------------------------------------
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
