Gilles Gouaillardet

tagging @reustle @liquidx to make sure you got that blocker issue.

sounds great! what about the Deceased patients? can we use the new 'Count' column? or do we still need one entry per patient?

Here is my counter offer: the CMake patch should be enhanced in order to also look for `ompi-c.pc` and friends.

For now, there are two main MPI implementations: `MPICH` and `Open MPI`, and each have several derivatives. So from a pragmatic point of view, this limits scalability to `2`. Note...

@rhc54 the issue involves PMIx From the logs, Open MPI configures the internal PMIx with ``` checking for subdir args... '--disable-man-pages' '--with-pmix-symbol-rename=OPAL_MCA_PMIX3X_' '--enable-embedded-mode' '--disable-pmix-timing' '--without-tests-examples' '--disable-pmix-binaries' '--disable-pmix-backward-compatibility' '--disable-visibility' '--enable-embedded-libevent' '--with-libevent-header=opal/mca/event/external/external.h'...

Thanks @jsquyres, that will do the trick. @rhc54 the root cause is in both Open MPI and PMIx, and I will backport the fix to PMIx once reviewed here. @leleliu008...

@rhc54 I will also issue the PR for PRRTE if needed.

Thanks @leleliu008 for the lengthy explanation. I updated the commit message of #9131 to credit you as LiuFuPin (I am not sure we can use kanji - nor other non...

Assuming there are more than 2 MPI tasks and you have enough nodes, what about ``` mpirun --rank-by socket ... ```

Can you try ```--rank-by node``` instead?