LorenaABarbaApr 29, 2023

Interesting report: "An evaluation of risks associated with relying on Fortran for mission critical codes for the next 15 years," Los Alamos National Lab. Finds that it's very likely they will be unable to staff #Fortran projects with top computer scientists/engineers, and even chance they can get computational scientists…

Link to PDF: https://permalink.lanl.gov/object/tr?what=info:lanl-repo/lareport/LA-UR-23-23992

[info:lanl-repo/lareport/LA-UR-23-23992] An evaluation of risks associated with relying on Fortran for mission critical codes for the next 15 years

danibanezMay 1, 2023
Show threadBalthasar Reuter
@labarba it has also sparked an interesting discussion among proponents of the language: https://fortran-lang.discourse.group/t/an-evaluation-of-risks-associated-with-relying-on-fortran-for-mission-critical-codes-for-the-next-15-years/5644
An evaluation of risks associated with relying on Fortran for mission critical codes for the next 15 years

https://permalink.lanl.gov/object/tr?what=info:lanl-repo/lareport/LA-UR-23-23992

Fortran Discourse
Apr 29, 2023 at 11:56pmWeb
Show threadarclightApr 30, 2023
@reuterbal @labarba Good discussion with some very knowledgeable people there. With internal mentorship and training and suitable incentives, I'm certain LANL could build and maintain the expertise they need. I don't believe anyone wants to pay for that. I understand their attraction to C++ but I find it is a far, far more difficult language to master, especially for subject matter experts vs CS grads with no domain knowledge.
Show threadJed BrownMay 1, 2023
@arclight @reuterbal @labarba I don't love C++, but it has orders of magnitude more investment in tooling and bugs are more rare despite the very complicated spec. I'd rather see more uptake of Rust, which has the industry commitment and fantastic tooling. Like it or not, Fortran rarely attracts that sort of language and tool building talent at any price, and the language semantics make it hard to provide a good experience.
Show threadBalthasar ReuterMay 1, 2023
@jedbrown @arclight @labarba 100% agree. The widening gap between the small amount of tooling for Fortran on the one side and a rich ecosystems on the side of C++, combined with a lack of vendor commitment to the language (e.g., only OpenMP by Intel/AMD) are making it hard already today. This can't be fixed in a few years and, moreover, any training that institutions can provide can only cover their own needs, not create the dynamics of a vivid ecosystem