| 5:02:20 AM PST - Thu, Feb 1st 2018 |
|
so is this a result of me using
ARMCI_NETWORK=MPI-PR
This
http://nwchemgit.github.io/index.php/Release66:ARMCI
says
"Therefore, when executing on a single node (i.e. the case of desktop execution) you would need to ask for n+1 processes; in other words, a serial execution would require the following mpirun invocation mpirun -np 2"
but both procs seem equally busy (though one is using a lot more memory than the other . . .)
Not entirely certain . . .
--
Jason
|