[petsc-dev] Using Hypre AMS and Standard Hypre AMG

[Ari Rappaport]

Hi all,

I am running a problem that requires both preconditioning for an elliptic problem,
as well as a problem stemming from Maxwell's equations. Currently, I am using PETSc's
bridge to Hypre for both, PCHYPRE.

The problem is that the choice of whether to use Hypre's special AMS solver is decided
at the call to PETScInitialize using a command line argument "-pc_hypre_type ams". 
I would like to be able to decide to use the ams space once, and then the standard AMG 
once, all in the course of a single run. Does PETSc support this capability somehow? 
Is it possible to call PETScInitialize multiple times safely perhaps?

Thank you,
Ari