Using coarrays on distributed architecture with hydra?

[paul_romano]

I'd like to compile a program with coarrays for a distributed memory architecture (Linux cluster consisting of 18 nodes each with 2 four-core processors). I see that the compiler uses Intel MPI as its backend which is based off on MPICH2. By default, this means that to run MPI processes, an mpd daemon (or ring of daemons) needs to be running on the nodes. When I set up such a ring using mpdboot, everything works fine. However, I'd like to use the Hydra process manager that is the new default process manager in MPICH2-1.3+. When I change the symbolic link for mpiexec to mpiexec.hydra instead of mpiexec.py, I receive the following error:

bash: /opt/intel/composerxe-2011.0.084/mpirt/bin/intel64/pmi_proxy: No such file or directory

Is there any way to get Intel MPI working with the Hydra process manager?

Best regards,
Paul

[Steven_L_Intel1 · Accepted solution]

My understanding is that Intel MPI does not yet support Hydra. You might ask in the Clusters section of the forum. But our coarray implementation is assuming "vanilla" Intel MPI at this time.