[mpiwg-languages] static datatypes lifetimes

[Ghosh, Sayan]

Good point to raise, Alfredo.
Tim – were you referring to this?


  *   Delete callback associated with communicators gets called when communicator is freed
  *   Consider finalize-delete-callback (this is what Alfredo is alluding to perhaps w.r.t datatype-attached-to-environment) – that seems to rely on MPI_COMM_SELF callback (freeing comm-self triggers callback)
--
Sayan Ghosh
Computer Scientist
Future Computing Technologies Group
Pacific Northwest National Laboratory
(509)-372-6323 • https://urldefense.us/v3/__https://sg0.github.io__;!!G_uCfscf7eWS!dDshRCeikLXOU8tk7BK4FrcfSKTeoJd4BHT-vBSD0yd_25QhX9DezifApWMlNnbtD1-LGkYDkcR-wbWl9odOPMR_YzXUmYnHr7UdKQ$ 


From: mpiwg-languages <mpiwg-languages-bounces at lists.mpi-forum.org> on behalf of Alfredo Correa via mpiwg-languages <mpiwg-languages at lists.mpi-forum.org>
Date: Friday, April 25, 2025 at 1:45 PM
To: mpiwg-languages at lists.mpi-forum.org <mpiwg-languages at lists.mpi-forum.org>, tim.uhl at kit.edu <tim.uhl at kit.edu>
Cc: Alfredo Correa <alfredo.correa at gmail.com>
Subject: [mpiwg-languages] static datatypes lifetimes
Check twice before you click! This email originated from outside PNNL.

Hi Tim, the other day you mention that you played with attaching datatypes (specifically static data types) as keyvalues. This is very interesting. The first thing I notice is that these datatypes will not be global, which is good. but it also
ZjQcmQRYFpfptBannerStart
This Message Is From an External Sender
This message came from outside your organization.

ZjQcmQRYFpfptBannerEnd
Hi Tim, the other day you mention that you played with attaching datatypes (specifically static data types) as keyvalues.

This is very interesting. The first thing I notice is that these datatypes will not be global, which is good. but it also means that they are attached to communicators, not the MPI environment. Am I correct?

This would solve the problem of freeing datatypes on program exit.

This is different from the idea of attaching it to environment that I was mentioning, but it is possible that attaching it to the communicator is actually a better idea. At the very least, because MPI has a mechanism to “attach” stuff to communicators, but not to the environment.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-languages/attachments/20250425/23695f89/attachment-0001.html>