Most code bases don't care about latest standards, rather quality of generated code.
Exactly, that is my point: so we are mainly still in the land of Fortran95, which is not designed for taking advantage of the latest features of the hardware.
Which doesn't change the current situation that commercial compilers still win, and NVidia/ARM contribution to LLVM comes from PGI experience.
gfortran also does not do GPGPU.
It was quite surprising to find out the lack of understanding from Khronos that CUDA supports Fortran out of the box, during OpenCL 3.0 Q&A session.
But then we are not talking about fortran anymore: I tried recently PGI and it is far behind than intel/gfortran with respect to the standard.
I can code with CUDA in many more languages than fortran. If I wanted to use gfortran, I just would make a wrapper to CUDA with C++.
Sure we are, OpenAAC, MPI and GPGPU support are a standard feature of any commercial Fortran compiler.
And yet none of those technologies is fortran-specific, nor mentioned on its standard, nor solve at tiny bit the problems the OP was specifying.
Try to give a C developer a C compiler without support for inline assembly to see how they will appreciate it.
Yet, inline assembly is not part of ISO C.
Or one that generates lousy machine without full advantage of the CPU vector units.
Yet, vector instructions are not part of ISO C.
I don't think commercial/open source is the key here: on the fortran side we have been long suffering from a lot of bugs/regressions with the current versions of intel fortran (with respect to the "latest" features like OOP) -- I would even say that they could be more than we are finding in gfortran.
An explanation could be that they are investing time in supporting their big customers that likely use ancient code bases.