Combine single_PAO_to_grid and single_PAO_to_grad

[tkoskela]

Closes #244

I've merged the two identical subroutines into PAO_or_gradPAO_to_grid. The subroutine called in the inner loop is passed in as a procedure argument through an interface. As a side effect I had to always pass direction into both subroutines to make them conform to the same interface, in evaluate_pao it's just a dummy argument that does nothing. I've updated all calls to PAO_or_gradPAO_to_grid, evaluate_pao and pao_elem_derivative_2 and added a bit of explanation in comments.

Very open to a more descriptive names for the subroutine and the interface

Gets rid of OpenMP overheads in single_PAO_to_grad we were seeing earlier

Previously:

This PR:

[tkoskela]

I don't see any noticeable performance impact of passing the subroutine as an argument instead of selecting it with an if statement inside the loop nest. I like not having an if statement at the bottom of the loop nest from a code readability point of view.

[davidbowler]

The name here isn't really quite right, but I can't think of an alternative! The previous names (single_PAO_to_grid and single_PAO_to_grad) but also misleading (the second should have been single_PAO_grad_to_grid).

I wonder if single_PAO_to_grid would be better though it might confusing historically

[tkoskela]

I don't understand the algorithm well enough to give the subroutine a descriptive name, so I'm really waiting for your input here 😅 If you think single_PAO_to_grid is good, let me know.

[davidbowler]

Possibly single_PAO_or_grad_PAO_to_grid which is at least descriptive! I don't think that we need the single part, so maybe PAO_or_grad_PAO_to_grid. @tsuyoshi38 and @ayakon do you have any thoughts?

[tsuyoshi38]

PAO_or_grad_PAO_to_grid sounds reasonable to me. (I would like to remove _ between grad and PAO, like PAO_or_gradPAO_to_grid

[tkoskela]

Technically this should be "any subroutine that satisfies the evaluate interface to grid" but I suppose we can name it after the current implementations. Would transform_to_grid be too generic? When you call the subroutine you'll see what you are transforming from the arguments, for example

call transform_to_grid(tmp_fn, pao_elem_derivative_2, dir1)

[davidbowler]

I think that would risk confusion as this routine is specifically for PAOs and their derivatives (there are other functions transformed to the grid). I've taken what @tsuyoshi38 and I agreed (PAO_or_gradPAO_to_grid) and pushed it.

[tkoskela]

Ok. Ready to merge then?

[ilectra]

Only minor aesthetic comments

[tkoskela]

Only minor aesthetic comments

Thanks! All done now