Artem Polyakov

Thanks, @ax3l! Very nice to have address sanitizer in place!

Thank you, Ralph.

Regarding the master. Looks like the following RETAIN is redundant: https://github.com/openpmix/openpmix/blob/master/src/client/pmix_client_get.c#L251 Removing it makes the master behave.

Thank you for the great summary! I'd like to point @samuelkgutierrez to few informational sources that might be useful to ramp up: * Our 2019 EuroMPI paper *A performance analysis...

**dstor limitations**: * Support for direct pointers in the shmem region: * Should be easy to extend it to support trivial data types like int, float, etc. * Complex datatypes...

At the moment, we are using serialized representation that is generated during modex operation. We just delay the unpacking on the server and place data into the shared memory. Client...

This PR depends on #11 to avoid conflicts.

Will rebase ASAP as indentation PR is in

@cchambreau, rebased

This should be a fairly easy one as well if I'm not missing something.