Standard conformance tables

Fortran standard conformance tables

The latest version of the tables are now available.

Here is a link to our information page

Here is a link to the standard conformance tables

6 Likes

Great! but I don’t understand why the Fortran 2008 conformance is based on GFortran 9.x (the 9.1 was released in May 2019) and the Fortran 2018 conformance is based on the still older GFortran 7.x ? It sounds weird…

Apologies for the inconsistency. When creating the 2018 table we started with gfortran 7.5, and I’m not sure why. Mea culpa. The vendors provide the data/updates and we incorporate the information into the tables.

4 Likes

GFortran 7.5 was also released in 2019, November:
https://gcc.gnu.org/gcc-7/

It’s quite old as we are now at Gfortran 12.2 (August 2022)
https://gcc.gnu.org/gcc-12/

We rely on the vendors provding the updates. We haven’t had any updates to the 2008 table for a while, and no updates to the 2018 table at all.

1 Like

Maybe you could simply add a line in the tables with the release date of the compilers. It would give a clue to the reader who is not familiar with certain compilers and don’t know what is their latest version.

Well I know for sure that select rank is not supported for gfortran < 10 but supported for gfortran >= 10.

You can see it mentioned here:
https://gcc.gnu.org/wiki/Fortran2018Status

And actually tested (by me) here:

So I guess you cannot rely on accurate updates from gfortran…

I checked that in your table select rank appears as not supported in gfortran, which surely matches the reported version and the β€œno updates received” statement.

Thank you for the updated tables.

I am also a bit confused about the PGI/NVIDIA compiler (2008 table contains PGI 18.1, 2018 table has Nvidia 16.4 instead), could you clarify that? I would expect here a recent version (22.x) of nvfortran.

Regarding the outdated data, maybe it would be easier to collect information if the browsable database discussed in Compiler support tables was implemented? In the end, we want to increase vendor engagement in the adoption of new language features, maybe this could become a PR thing where we use vendor inputs to rank compilers?

Thanks for brining this to our attention.The PGI/Nvidia issue goes back to the original version of these tables that appeared in 2020 in Fortran Forum.

We haven’t had any updates since then from Nvidia so we haven’t changed the data in the tables.

We update the tables when the vendors provide us with the information.

At the moment we don’t have any plans to implement a browsable on line version of the data.

If some wants to do that we have no issue with that.

Ian

1 Like

@cmaapic sir, I am willing to maybe digitalise or help to Bring an online searchable version of this data in my free time, how to go forward in this direction ( if you could please share your vision regarding this) ?

Thanks and Regards,
Henil

6 Likes

Thanks @cmaapic for sharing it!

I can provide the document in PDF, XML or HTML format. I don’t have any idea about how to provide an on line searchable version. Maybe some one else on the list can help here.

2 Likes

@cmaapic Ian, your table is very helpful for the community. It’s important to have such a complete catalog of feature support in compilers. Thank you.
Fortunately or unfortunately here at Intel we just released ifort 2021.7 and ifx 2022.2. Terrible timing on our part, apologies.
IFX and IFORT should now be equal wrt Standards support through 2018 with the exception of the 2018 REDUCE intrinsic, which is on our to do list for the next release for IFX.
We have updated our IFX Language and OpenMP support page
At this point both compilers should implement 2008, and 2018 save for REDUCE missing in IFX. Thanks again for this valuable document!
Ron
#IAmIntel

4 Likes

Does this imply that (almost) full F2018 features are now supported on the GPU? Does ifx allow to run arbitrary coarray code on GPUs?

C++ has a great standard conformance table:
https://en.cppreference.com/w/cpp/compiler_support

1 Like

@fortran4r it would be awesome if we could have such a table at fortran-lang.org. It should be automatically constructed from Fortran test files that must compile and run without any runtime errors.

3 Likes

The link is broken, but for the sake of others interested in details of the latest ifort/ifx release, here is the OneAPI release note.
Particularly eye-catching for me is β€œCoarrays, including Fortran 2018 teams and events, are now fully supported. (F2008 and F2018 support)”.
The impressive progress on ifx deserves huge applause.

4 Likes

Fortran used to have such nice standard conformance tables until very recently. I cannot find them anymore on the web.

ACM Fortran Forum has details of standard conformance tables. Here is a link. https://dl.acm.org/institution/60272075. The first article appeared in 2007. Here is a link https://dl.acm.org/doi/10.1145/1243413.1243414. The last appeared in February 2021. https://dl.acm.org/doi/10.1145/3432987.3432991 The first post in this thread has details of the latest version Jane and I have worked on, which now addresses the Fortran 2008 and Fortran 2018 standards.

2 Likes