Cloud Hosting

A new report from Los Alamos National Lab sounds alarms over the declining number of Fortran programmers, the shrinking number of efforts to teach Fortran, and the reduced appetite of scientists and developers to learn Fortran. Developed by IBM in the mid 1950s, Fortran was a foundational programming language for scientific computing. In recent years it has been largely overtaken by modern programming languages such C++ and Python.

This latest report (An evaluation of risks associated with relying on Fortran for mission critical codes for the next 15 years) paints worrisome picture.

We judge it is very likely that we will be unable to staff Fortran projects with top-rate computer scientists and computer engineers, and that there is an even chance we will be unable to staff Fortran projects with top-rate computational scientists and physicists, write the LANL researchers, Galen Shipman and Timothy Randles. They offer the following supporting bullets:

Looking at modern HPC hardware infrastructure, the report says, We judge it is very unlikely that codes that rely on Fortran will have poor performance on future CPU technologies, it is likely that codes that rely on Fortran will have poor performance for GPUs, and it is very likely that Fortran will preclude effective use of important advances in computing technology. (emphasis added)

Zeroing in on vendor support, Shipman and Randles write, The vendor ecosystem of Fortran compilers is worrying. Intel and GCC communities have the most robust Fortran compilers for modern Fortran (Fortran 2008) on CPU technologies but have less mature support for GPU technologies. Nvidia has good support for GPU technologies but lacks support for modern Fortran needed by LANL.

Open-source efforts around an LLVM compiler for Fortran, known as Flang, are inadequate to meet either requirement (robust support for modern Fortran and GPU technologies). Complicating things further, there are competing Fortran technologies for GPUs including standards such as OpenACC and OpenMP and vendor proprietary technologies such as Cuda Fortran. While similar diversity exists for other languages (such as C++) there are no infrastructures for portability like Raja and Kokkos for Fortran.

Ouch.

The relatively short report is best read directly. On balance, its message isnt new but part of long-term chorus of worry about Fortrans decline and the impact of that decline on HPC and legacy scientific codes.

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

See the original post:

LANL Report Looming Fortran Talent Scarcity is Threatening - HPCwire