[mpiwg-tools] Fwd: [mpiwg-sessions] notes from 7/29/19 webex

[Marc-André Hermanns]

Dear all,

attached you find Howard's notes on a call we just had among the
people involved with PRs currently in need of MPI_Info before init
(MPI_T Events, Session, FT).

Cheers,
Marc-Andre


-------- Forwarded Message --------
Subject: 	[mpiwg-sessions] notes from 7/29/19 webex
Date: 	Mon, 29 Jul 2019 18:29:45 +0000
From: 	Pritchard Jr., Howard via mpiwg-sessions
<mpiwg-sessions at lists.mpi-forum.org>
Reply-To: 	MPI Sessions working group <mpiwg-sessions at lists.mpi-forum.org>
To: 	MPI Sessions working group <mpiwg-sessions at lists.mpi-forum.org>
CC: 	Pritchard Jr., Howard <howardp at lanl.gov>



Hello Folks

 

Some notes from the 7/29/19 webex.

 

Attending: Dan Holmes,  Wesley Bland, Aurelien Bouteiller, Marc-Andre
Hermanns, Howard Pritchard

 

Dan turned on recording of the meeting.

 

Dan pointed out a conflict about info keys – namely how they will be
accessed before MPI initialization
and after MPI finalization.  Dan and Mark-Andre had discussed at ISC
and were of the opinion that
changes for MPI info keys should go into the sessions (PR 55)
proposal.  Sessions is being considered as
a gating factor for 4.0.  We need to figure out what is the difference
between the info object changes
in the. 

 

Aurelien says all the FT WG is wanting is to have MPI_INFO_ENV
available before MPI initialization so
the app can figure out what error handler has been set.  Tools and
Sessions don’t really care so much
about this being available or not before MPI initialization.  Dan
discusses compromise that Sessions
has taken.  Wesley says FT needs MPI_Error_class and MPI_Error_string
to be invocable
prior to MPI initialization as well.  Aurelien thinks this is in PR 50.

 

Dan thinks the issue of argc/argv and how to fill in MPI_INFO_ENV is
the key question.

 

Aurelien thinks maybe add a new function to set MPI_INFO_ENV?    Maybe
a new ticket to handle
this.    FT group will work on this.

 

Dan says we need to go through discrepancies between tickets. 
Sessions will need to pull in some of
the FT adds to callable-prior-to-MPI Init to the table in Chapter 10
of PR 55.

 

Mark-Andre raised some concerns about MPI_Info_get and error cases,
but we think this will be
covered by the setting-default-error handler before MPI initialization
approach. 

 

Howard

 

-- 

 

Howard Pritchard

HPC-ENV

Los Alamos National Laboratory

 


-- 
Dr. rer. nat. Marc-André Hermanns

IT Center
Group: High Performance Computing
Division: Computational Science and Engineering
RWTH Aachen University
Seffenter Weg 23
52074 Aachen
Phone: +49 241 80-24381
hermanns at itc.rwth-aachen.de
www.itc.rwth-aachen.de
-------------- next part --------------
_______________________________________________
mpiwg-sessions mailing list
mpiwg-sessions at lists.mpi-forum.org
https://lists.mpi-forum.org/mailman/listinfo/mpiwg-sessions
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5336 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-tools/attachments/20190729/ed1a6094/attachment.p7s>