I wanted to call your attention to an on-going series of videos on the
YouTube channel Dave's garage:
https://www.youtube.com/watch?v=pv4Yq35Chx0
This all began when the author of the Youtube channel recounted that he
implemented a prime number sieve as an exercise in his school years in
the '80s, and now wanted to run exactly the same benchmark using modern
languages (initially, C# and Python).
This proved very popular, and watchers have submitted their
implementations in lots of languages (61 and counting), including Fortran.
Specifically, the goal is to implement a prime number sieve for primes
up to a million, closely following the algorithm in a reference
implementation (in C#).
In every episode of the series the author consider a couple of
implementations in similar languages (e.g., C# and Java), discusses a
bit the languages and then shows the results of the benchmarks.
There is a Github repository with all the implementations
https://github.com/PlummersSoftwareLLC/Primes
At the moment there are two Fortran implementations, one Fortran 2003
and 2008.
https://github.com/PlummersSoftwareLLC/Primes/tree/drag-race/PrimeFortran
I think it's be great if some of the Fortran experts had a look at those
implementations and, perhaps, improved them if possible.
I think if Fortran did well this would be good publicity for the general
public. I found some benchmarks (for older implementations, as they are
updated all the time) here:
https://www.reddit.com/r/fsharp/comments/oezbnv/daves_garage_prime_numbers_laguages_benchmark/
The list above indicates that the fastest implementations are in C and
zig (a language I'd never heard of), with the Fortran implementation far
behind (although many of those results look very strange, so take them
with a pinch of salt).
Lorenzo