Should all 'parallelization' related arguments be converted to `parallel_args'?

[jgostick]

We have some inconsistency in the way different function accept/expect parallelization arguments. Some accept a dict called parallel, others accept divs and then look at the settings to see how many cores to uses, etc. Both are not great.

I am thinking parallel_kw and it takes a dict of keyword arguments and passes them on as needed.

[ma-sadeghi]

Good point.

What if we just didn’t accept divs or any other parallelization-related kwargs? Since those are really part of global settings (which is why they’re stored in settings), they don’t belong in the function signature.

Then again, divs might be different since it helps avoid memory overflow on low-memory machines, so maybe we should keep it. But I still think the rest of the parallelization arguments don’t belong in the function signature.