MPI_STATUS_SIZE, MPI_SOURCE, MPI_TAG, MPI_ERROR not defined in C

[jsquyres]

Problem

The constants MPI_STATUS_SIZE, MPI_SOURCE, MPI_TAG, and MPI_ERROR are specifically not defined in C.

Proposal

Add MPI_F_STATUS_SIZE, MPI_F_SOURCE, MPI_F_TAG, and MPI_F_ERROR constants in C. The F in the names emphasizes that these are equivalent values to the already-defined-in-Fortran constants.

Note that the _SOURCE, _TAG, and _ERROR constants are 1 less than their Fortran counterparts (since they are offsets into an array, and Fortran and C differ in their handling of array offsets).

Changes to the Text

1. Add MPI_F_STATUS_SIZE up in chapter 2/terms and conditions
2. Add the other MPI_F_ constants in chapter 18/language bindings

Corresponding PRs:

• MPI 3.x: https://github.com/mpi-forum/mpi-standard/pull/271
  • PDF: mpi-report-pr271-3449c7f2d948906ef6d66478974dec55e0b541c5.pdf
  • Changes:
    • p15:37 (terms and conventions, listing of constants)
    • p688:30-34 (Fortran bindings/status section)
    • p704:6, 14-19 (update constant tables in A.1.1)
    • p837:31-33 (changelog)
• MPI 4.x: https://github.com/mpi-forum/mpi-standard/pull/270
  • PDF: mpi-report-pr270-266490d597c55671f7e6bb78380f2bcdae97c1bc.pdf
  • Changes:
    • p15:37 (terms and conventions, listing of constants)
    • p758:33-37 (Fortran bindings/status section)
    • p774:6, 14-19 (updated constant tables in A.1.1)
    • p915:31-33 (changelog)

Impact on Implementations

Add MPI_F_STATUS_SIZE, MPI_F_SOURCE, MPI_F_TAG, and MPI_F_ERROR constants in mpi.h.

Impact on Users

Users can allocate INTEGER-style Fortran statuses in C and properly read their values.

References

https://lists.mpi-forum.org/pipermail/mpiwg-fortran/2020-September/003584.html

[RolfRabenseifner]

As far as I can see, needed and a perfect solution for this bug in MPI-3.1.

[wesbland]

This passed an errata vote on 2020-09-28.

https://www.mpi-forum.org/meetings/2020/09/votes