Stefano's fixes and improvements

[stefanozampini]

I will use this PR to provide my feedback to the project

First thing, you do not need to test for HYPRE support at configure time, PETSc provides a macro for it (PETSC_HAVE_HYPRE). In general, PETSC_HAVE_XXX indicates PETSc has been configured with support for XXX package. Similarly, PETSC_USE_COMPLEX indicates PETSc has been configure with PetscScalar == complex_number.
Do you want me to fix also the complex thing? Note that PETSC_HAVE_COMPLEX, just indicates that std::Complex is available to PETSc, see here https://gitlab.com/petsc/petsc/-/blob/master/include/petscsystypes.h#L208

[stefanozampini]

Having few examples of usage from C++ would be nice, since the interface library is callable from C++. I can only find python examples in the project directory.

[stefanozampini]

Other initial comments:

• what is the minimal version of PETSc you want to support?

• I think you should not expose petsc.h in public headers. You can include petscsystypes.h for basic datatypes (eg PetscInt, PetscScalar, etc), and just forward declare the PETSc classes that are visible through the interface in src/typedefs.hpp, see e.g. https://github.com/mfem/mfem/blob/master/linalg/petsc.hpp#L42

[stefanozampini]

Pushed some more comments directly in the code 6e96003

[LukasKogler]

Thank you for all your suggestions!

I merged your commit for the HYPRE support, this is really much cleaner!

what is the minimal version of PETSc you want to support?

I am not sure, we are discussing this. Do you have a suggestion? Do PETSc users tend to use older PETSc versions, or do they generally keep up with the releases?

Do you want me to fix also the complex thing?

I did it myself, thank you for the hint about PETSC_USE_COMPLEX!

I think you should not expose petsc.h in public headers

I changed this the way you suggested.

Having few examples of usage from C++ would be nice

I added a simple C++ example.

As for the comments in the code:

• I still have trouble interfacing with a 64 bit index PETSc installation (linking issues).
• The petsc_ops.hpp file is not used, I just forgot to remove it. Sorry for the inconvenience.
• I put in the code you suggested in petsc_snes.cpp
• SBAIJ support is indeed still missing
• Using AIJ/MPIAIJ instead of BAIJ/MPIBAIJ per default might be a good idea. I have to write the functions to directly assemble AIJ matrices from our block-matrix format.
• I am still experimenting with PCFIELDSPLIT (and MATNEST). Thanks for the warning!

Thanks again for your feedback, the inteface already looks much neater!

Best,
Lukas