Apologies in advance that this might be a naive question.
I'm working under Andy Leigh Brown at Edinburgh, with HyPhy installed on Eddie (I believe Sergei did this himself a few years ago).
The original version, in Sergei's space on Eddie, seems to have problems running GADatedClock.bf with multiple nodes using MPI (it will run if I mpirun HYPHYMPI but ask only for 1 node). This version is from 2009 and I found quite a few differences from the current code in the file where the error seemed to be originating, so I have compiled and installed the latest HyPhy MPI code onto Eddie.
The new version seems to be running without errors. However, in the old version, HyPhy printed out a message to each node (in the message.log.mpinode files):
Quote:"[MPI] Entered mpiNormalLoop"
And also often printed out:
Quote:Row and/or column partition is empty. All the data will be used by default.
So I knew that HyPhy seemed to be successfully accessing each node. However, on the new version, I get no such messages in the message.log.mpinode files.
Is this simply how it is now, or is HyPhy not using the nodes for some reason? I have run the mpiTest.sh file which I found on this forum, and that seems to run sucessfully. Is there any way I can know for sure that HyPhy is being able to use the other nodes (force it to print something, perhaps?)?
Apologies again if there is an obvious way to figure this out!
Thank you,
Emma