Deprecate MPI_HOST

[jdinan]

Summary

The MPI_HOST attribute on MPI_COMM_WORLD is ambiguous and no longer relevant for MPI implementations. This was reviewed at the September, 2015 meeting and there was consensus that it should be deprecated.

Proposed Change

Page 228, lines 3-4: Remove or reword sentence containing "host process"

Page 335, line 4: Remove line beginning with "MPI_HOST"

Page 335, lines 27-33: Remove "Host Rank" subsection.

Page 675, line 47: Remove "MPI_HOST" from "Environmental inquiry keys" table

[abouteiller]

This issue has been duplicated by #201

[jaegerj]

Annotated MPI 4.0 pdf with red changebars.
Changes are listed on first page.
mpi40-report-deprecate-MPI_HOST.pdf

And annotated MPI 3.1 pdf for comparison
mpi31-report_depracate_MPI_HOST.pdf

[RolfRabenseifner]

A deprecated feature is still available and must be therefore fully explained. The following text from MPI-3 and MPI-4 is missing in the proposed pdf:

The value returned for MPI_HOST gets the rank of the HOST process in the group associated
with communicator MPI_COMM_WORLD, if there is such. MPI_PROC_NULL is returned if
there is no host. MPI does not specify what it means for a process to be a HOST, nor does
it requires that a HOST exists.
The attribute MPI_HOST has the same value on all processes of MPI_COMM_WORLD.

Also in Annex A.1, the constant MPI_HOST must not be removed.

Recommendation: Add only comments that MPI_Host is deprecated since MPI-4.x in the appropriate section and A.1.1. The additional text in the Deprecated chapter and the change-log is then enough.

And it must be listed in Table 2.1 also missing in PR https://github.com/mpi-forum/mpi-standard/pull/221

[jaegerj]

Hi Rolf,

yes, I did not take the time to correct this PR yet. This is not intended to 4.0 anymore. I plan to update the PR after the next meeting.

[RolfRabenseifner]

If a feature is such integrated in other parts of the standard (as in the past the C++ interface, see MPI-2.2), then the best way may be to keep the existing text, i.e., to start with a new PR from final MPI-4.0, and to add only deprecation remarks at the definition of MPI_HOST, at the repeating in A.1.1, and to add a list entry in Table 2.1, a paragraph to the Deprecated chapter, and an entry in the change-log.
I saw that you flagged with MPI-4.1.

[wesbland]

You don't need to close the PR. Just push updates to it...

[wesbland]

This wasn't read at the December 2022 meeting. The last opportunity for MPI 4.1 is to have it ready at the March 2022 meeting (and it needs to "pass" the reading).

@jaegerj Are you (or is someone else) planning to push this forward or should we move it out of the plan for MPI 4.1?

[schulzm]

This is scheduled for discussion in the virtual meeting on January 11th

[RolfRabenseifner]

@jaegerj When you announce the 1st vote, you may mention that there were a view editorial changes since the reading.

[wesbland]

There's two PRs that point back to this issue. Which one is the one that was voted on?

[jaegerj]

The most recent one: https://github.com/mpi-forum/mpi-standard/pull/790

[mpiforumbot]

This passed a no-no vote.

Yes	No	Abstain
31	0	0

[mpiforumbot]

This passed a 1st vote.

Yes	No	Abstain
29	0	2

[wesbland]

This passed a 2nd vote.

Yes	No	Abstain
28	0	4