Support Fortran 2018 ISO_Fortran_binding.h

[rouson]

The draft Fortran 2015 standard defines a new C header file described as follows in Subclause 18.5.1:

"The source file ISO_Fortran_binding.h shall contain the C structure definitions, typedef declarations, macro definitions, and function prototypes specified in subclauses 18.5.2 to 18.5.5. The definitions and declarations in ISO_Fortran_binding.h can be used by a C function to interpret and manipulate a C descriptor. These provide a means to specify a C prototype that interoperates with a Fortran interface that has a non-interoperable dummy variable (18.3.7)."

Gfortran developer Paul Richard Thomas recently suggested that this header can be implemented without touching other parts of the gfortran compiler, which is an intriguing possibility. Rather than changing the gfortran descriptors, for example, the header can make use of functions that take fortran's native descriptors as input and return the standard descriptors as output. This issue is being created for discussion by those interested in this feature and/or in contributing to this feature.

[rouson]

@zbeekman this could also be a good project for someone interested in learning more C. ;) For now, I'm hoping to find some time to work on this myself and @sfilipone (not sure why tagging him doesn't work here) has offered to help at least with testing at a minimum.

[zbeekman]

strange that @sfilippone doesn't work... did he change his account or something?

EDIT: oh, nm, it looks like @rouson had a typo.

[rouson]

I just created an issue-#6-iso-fortran-binding-h branch with a start on iso_fortran_binding.h. It's incomplete and probably even incorrect in a few places, but it would be good to get some initial feedback from @jerryd regarding style, e.g., regarding the format and verbosity of the comments.

[sfilippone]

Didn't change anything. What is not working? S.

…

On 20 Oct 2017 21:31, "zbeekman" ***@***.***> wrote: strange that @sfilippone <https://github.com/sfilippone> doesn't work... did he change his account or something? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#6 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/APX-TxV5JP6t4ww6lNLmQDxW4IqKMDsHks5suQMIgaJpZM4QBLoM> .

[jerryd]

Need s space between function names and first left paren. Tabbing should be two space indents, 8 spaces converted to tabs. (My editor just does that for me.) Lines should not exceed 80 characters and manually wrapped and indented appropriatly as needed. Comments end with a period followed by two spaces before the closing */

I will clone this down to my workstation and edit the file if you would like. Let me know.

[rouson]

Thanks, Jerry. Yes, please clone away! Either push to this branch or create a new branch and submit a pull request. Best Regards,

…

_______________________ Damian Rouson, Ph.D., P.E. President, Sourcery Institute www.sourceryinstitute.org +1-510-600-2992 (mobile)

On October 21, 2017 at 10:56:28 AM, Jerry DeLisle ***@***.***) wrote: Need s space between function names and first left paren. Tabbing should be two space indents, 8 spaces converted to tabs. (My editor just does that for me.) Lines should not exceed 80 characters and manually wrapped and indented appropriatly as needed. Comments end with a period followed by two spaces before the closing */ I will clone this down to my workstation and edit the file if you would like. Let me know. — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#6 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AMgGhLYsXPh-XbxJmJiUp7zrz69SloG3ks5sujBMgaJpZM4QBLoM> .

[zbeekman]

Brief recap of pertinent notes from call on 03/30/2018:

It would be useful to have an example of using the mpi-f08 module @scrasmussen has agreed to kindly help us by providing simple examples with send and receives, or some form of communication.

[scrasmussen]

@rouson, in f08_mpi.tar.gz is the send receive example using mpi_f08. The mpi_f08 module doesn't change much, just a few types.

Earlier in the call I remembered a few things incorrectly, I believe Intel's compiler has ISO_Fortran_binding.h, and as you can see, GNU doesn't. OpenMPI has a mpi_f08 module. Though I see it in MPICH's codebase, the MPICH version I installed with OpenCoarray's install.sh didn't seem to have it. I'm looking further into this.

If you have any questions please ask! I've spent a fair deal of time looking at these issues since I was writing a source to source compiler to generate wrappers to deal with C and Fortran interoperability for MPI.

[rouson]

@soren Thanks for attaching the mpi_f08 example. One confusing thing is the reference to MPI_COMM_WORLD followed by the subsequent reference to the declared bu uninitialized comm. Let's talk about this example during Friday's call.

[soren]

Hey, @rouson, you wanted to ping @scrasmussen, not me ;-)

[rouson]

@jerryd I just pushed a commit to the issue-#6-iso-fortran-binding-h branch of this fork putting the current versions of ISO_Fortran_binding.h and ISO_Fortran_binding.c in the libgfortran/runtime subdirectory as we discussed. If you have a moment to edit the GCC build system on the same branch or submit a pull request against the branch to incorporate the functions into libgfortran and install the header file, I would greatly appreciate it. That would make it easiest to start testing whether MPICH will find and recognize the file and therefore build the mpi_f08 module that the MPI standard describes. Currently, MPICH won't build mpi_f08 because it doesn't find ISO_Fortran_binding.h. Thanks for any assistance you can offer!