I think coarrays have a great potential for the future, but they are unfortunately not ready yet for the kinds of production codes that LANL has (Coarrays: Not ready for prime time).
I must strongly object how the paper is criticized here in this thread by many. We are losing all credibility if we are only addressing the way the paper was written, and not taking the actual problems with Fortran seriously.
Is it not our point to try convince big players to keep using Fortran? The answer should be yes!
So let’s address the actual issues that they have.
Just read the page 6 of the report to get started on the serious issues that Fortran has, I just picked a few:
-
As a specific example, efforts to port xRAGE to the PGI Fortran compiler (now rebranded as the nvfortran compiler) have been ongoing since April of 2019 and xRAGE is still unable to be built and run with this compiler. It is important to note that nvfortran still does not support the Fortran 2008 standard, a full 15 years after its ratification by the standards committee.
-
Multiple competing standards for Fortran-based GPU programming with varied levels of robustness and support exist today (Fortran OpenMP Target offload and OpenACC). Neither of these technologies is robustly supported on the AMD GPU (MI250) today.
Do we all agree that these are real issues that we need to fix, or at least provide a plan how they will be fixed and when?