Samuel K. Gutiérrez

Thank you @rhc54 and @artpol84 for the nice description of the problem and requirements. > It is relatively straightforward to achieve the second requirement when referencing simple values (e.g., int...

Thank you, @artpol84. I will look around the codebase to better understand the details, but that makes sense.

Regarding hwloc's symmetric memory mapping trick: [here](https://github.com/open-mpi/hwloc/blob/ad1777f06d448edfbca90c99364ef12852e73262/tests/hwloc/shmem.c#L93-L127) is the only place that I found use of that methodology. I wonder if this trick is used only for testing?

To keep track of what is learned along the way. When enabled in a particular way, will Linux ASLR render the symmetric virtual address space trick unusable in practice? One...

@rhc54 Interesting. Could you or @bgoglin please point me to that code in hwloc? The only source I could find that attempts this trick is cited above, and it looks...

> I think I understand your confusion here. HWLOC provides the _tools_ for doing it, but it is up to the PMIx server to actually implement the key parts. Thank...

Perhaps we are talking about different topics here. I was under the impression that this thread was specific to how to deal with the shared-memory key-value service.

> > I was under the impression that this thread was specific to how to deal with the shared-memory key-value service. > > It is - but you were asking...

Thank you, @bgoglin and @rhc54. To summarize: * We do not expect Linux ASLR to cause issues in practice. * hwloc and Open MPI have used symmetric memory mapping successfully...