[mpiwg-p2p] Message matching for tools

[Marc-Andre Hermanns]

Hi George,

sorry, was on leave for the last couple of days due to the "street
carnival"
(https://en.wikipedia.org/wiki/Carnival_in_Germany,_Switzerland_and_Austria#Rhineland)

> Assuming that the usual precautions are taken, aka. tools and
> libraries create their own communicators, I fail to see how the user
> hints on it's communicators can have any impact on libraries/tools ?

As Jim already pointed out in his reply, the problem arises when we
try to match messages solely on the envelope information. With this
assertion active, correct message matching in a replay as done by
Scalasca cannot be guaranteed, and neither can the wrong matching be
detected.

During the course of this discussion we agreed that this is not a
concern for the P2P working group, but the tools working group will
have to come up with something that mitigates the problem.

I don't know, if we need to modify the "Advice to users", though.
Because, once we found a way around the problem, we would could drop
the addition again. Thus, it would only be a temporary addition to the
standard, which I don't like very much.

My idea for an interim solution would be that the measurement tool
checks for the assertion and then points the user to additional
(non-standard) info hints---ignored by the MPI but seen by the
tool---that provide further guarantees for the tool.

Something along the lines of "scorep_deterministic_matching =
true/false". The tool would document this accordingly.

What do you think?

Cheers,
Marc-Andre

-- 
Marc-Andre Hermanns
Jülich Aachen Research Alliance,
High Performance Computing (JARA-HPC)
Jülich Supercomputing Centre (JSC)

Schinkelstrasse 2
52062 Aachen
Germany

Phone: +49 2461 61 2509 | +49 241 80 24381
Fax: +49 2461 80 6 99753
www.jara.org/jara-hpc
email: hermanns at jara.rwth-aachen.de

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4899 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-p2p/attachments/20160210/2e78bf74/attachment-0001.bin>