GitLab

While writing an adapter for the solver factory, I stumbled over the rather inconsistent naming of the parameters used for the solver factory. It is a pity that nobody looked at this before merging and that this is now part of a stable release.

We might want to change this ASAP as otherwise this might last forever.

It seems like the parameter names taken are just the names of the constructor parameters. Depending on who wrote the code, these are a bit too different or rather unexplicit:

• "verbose" is used for the solver, while the AMG uses "verbosity" (these should be the same to avoid confusion)
• "mmax" for the number of vectors stored in FCG is not very explicit
• "n" for the order of the ILU decomposition might still be ok, but order seems better

My questions would be:

• Shouldn't we settle on either verbose or verbosity?
• Shouldn't we use expressive parameter names (maxOrthogonalizationVectors instead of mmax
• Should we prefix preconditioner specific parameters with the the preconditioner "n"-> "iluN" or "iluOrder"?