Fortran Monthly Call: August 2021

[certik]

@zoziha

Yes, I hope that many implementations, such as the function sin , can be applied to arbitrary kind and type with only one template. It seems that kind (precision) is more important to me, after all, real(4/8/10/16) has four kinds of precisions, writing four copies of the sin function does not seem to conform to common sense. (or something like Function kind specified by optional input · Issue #91 · j3-fortran/fortran_proposals (github.com))

Or this: Allow an intent(out) argument to adopt the same kind as an input(in) argument · Issue #128 · j3-fortran/fortran_proposals · GitHub

@zoziha

Isn’t this generic content?

About any rank , we have elemental label and will have dimension(..) .

Yes it is in my mind, and I would even say this is the most important case. But the existing generic subgroup (I am part of it) has these use cases: https://github.com/j3-fortran/generics/blob/main/J3-Papers/Generics_Use_Cases.txt, and I need to make sure it prioritizes the above use case more.

[Arjen]

Note that there are (somewhat cumbersome) techniques to achieve support for multiple kinds with just one source version. If you want more variations, like various ranks, then either elemental routines will help or you need preprocessing like with fypp. But there is no need perse to keep multiple copies around merely for the sake of various kinds.