[mpich-discuss] mpix_comm_shrink hangs for more than 32 ranks

Atis Degro atisdegro at gmail.com
Fri Jul 12 05:46:15 CDT 2019


Dear MPICH community,

I am working on fault tolerant code
and using the mpix_comm_shrink call
to exclude the failed processes from the
communicator.
Everything seems to be working fine if
the size of the communicator is 32 or smaller.
If I increase the size however, the mpix_comm_shrink
call fails (hangs).

Is this a known issue and is there any workaround to
enable this call for larger communicators?

Thank you!

Atis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpich.org/pipermail/discuss/attachments/20190712/b75d0cd2/attachment.html>


More information about the discuss mailing list