[mpiwg-sessions] Fwd: World/session model

[HOLMES Daniel]

Hi Victor,

Thanks for the comments. We’ll discuss them within the WG in the new year.

Cheers,
Dan.
—
Dr Daniel Holmes PhD
Architect (HPC Research)
d.holmes at epcc.ed.ac.uk<mailto:d.holmes at epcc.ed.ac.uk>
Phone: +44 (0) 131 651 3465
Mobile: +44 (0) 7940 524 088
Address: Room 2.09, Bayes Centre, 47 Potterrow, Central Area, Edinburgh, EH8 9BT
—
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
—

Begin forwarded message:

From: Victor Eijkhout <eijkhout at tacc.utexas.edu<mailto:eijkhout at tacc.utexas.edu>>
Subject: Re: World/session model
Date: 18 December 2020 at 15:58:20 GMT
To: HOLMES Daniel <d.holmes at epcc.ed.ac.uk<mailto:d.holmes at epcc.ed.ac.uk>>

This email was sent to you by someone outside the University.
You should only click on links or attachments if you are certain that the email is genuine and the content is safe.

My cc to wg-sessions was rejected. I’m resending this in case it didn’t reach you.
%%%%%%%%%%%%%%%%%

If a program uses both the Sessions model and the World model


Thanks. At first I thought the two models were mutually exclusive but I have since found statements that they can be used together.

But then I find statements such as

7.2.4 When using the Sessions Model (Section 11.3) for initialization of MPI
re- sources, MPI_COMM_WORLD and MPI_COMM_SELF are not valid for use as
a communica- tor.

….a little very confusing. To me this sentence says that if I do anything with sessions, the use of MPI_COMM_WORLD is 100 percent ruled out. You mean to say: is ruled out for session-related activities. It certainly confused me.

Likewise, in 11.3 there is a very absolute sounding statement “MPI_COMM_WORLD is not valid for use as a communicator.” I’m sure that this sentence comes with an implicit qualification that it applies to using the Session model, but I’m not quite sure how.

While I’m nitpicking, 11.2.3 has an ambiguous sentence:

“For a library using got the World Model, it needs to know if MPI has been initialized using MPI_INIT  or MPI_INIT_THREAD”.

The wrong interpretation is “needs to know which of MPI_INIT / MPI_INIT_THREAD” has been used.

The correct interpretation (which really took me a minute) is “needs to know if MPI has been initialized using _either_of_ MPI_INIT / MPI_INIT_THREAD”

(Oh, and I think “For a library, it needs to know” is not grammatical. Either “A library needs to know” or “For a library, it may be necessary to know”)

Victor.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-sessions/attachments/20201218/c2473990/attachment-0001.html>