[petsc-dev] de-global variablelizing PETSc

[Karl Rupp]

Hey,

 >> Both a) and b) don't seem attractive to me, so I'd rather *not*
>> initialize everything in there.
>
> Agreed, I'm all for reducing how much is done in PetscInitialize.  It
> could be as bare-bones as taking command line arguments, malloc, and the
> profiling clock.  But I think it would be difficult to remove it
> entirely without cluttering user code with additional calls.  Ideally it
> could be simplified to a short sequence of function calls that the
> advanced user could call themselves.

Yep, that's how I envision this as well. As long as we don't have good 
(non-academic) reasons for removing PetscInitialize, I'd prefer to keep 
it around in order to have a well-defined entry-point for our users.

Best regards,
Karli