| 6:37:23 AM PST - Tue, Jan 12th 2016 |
|
Is this your entire error or was there a separate file for stderr? Ask ORNL for details as to why this crash occurred.
8000 is way too much for <1000 orbitals with MP2. I think you can run this on a single node workstation if you are patient.
I don't know why it fails, but you could try building NWChem with ARMCI-MPI instead. It is much more stable than Comex in my experience. See https://wiki.mpich.org/armci-mpi/index.php/Main_Page.
The other option, which is probably similar than ARMCI-MPI, is to use ARMCI_NETWORK=MPI-PR. This is what NWChem uses at NERSC and it crashes far less often than the Comex build.
|