mpirun needs "--bind-to none" specified if used for multiple executions in parallel

[cchall]

There is significant slowdown observed that scales up as the number of simultaneous executions of mpirun.
Adding --bind-to none see seems to mostly resolve the issue.

• Check if this issue exists as you add more cores to each execution
• Look at MPI specific executor settings for rsopt. Would probably want to only use this for openmpi. I'm not aware of any similar problem ever occurring on MPICH systems.
• Could be motivation to revisit a true Serial executor. I think there was talk of putting this into libEnsemble. Though it is difficult since we may want to accommodate a serial and MPI executor simultaneously.

[robnagler]

Sirepo always passes --bind-to-none. I don't think we've had issues with mpich.

[cchall]

Thanks @robnagler thats good to know