Something wrong with parallel output?

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
bourgeois
Posts: 3
Joined: Tue Mar 16, 2010 10:43 am
Location: Laboratoire de Physique des Solides, CNRS Orsay (France)

Something wrong with parallel output?

Post by bourgeois » Thu Apr 15, 2010 5:25 pm

Hi everyone,

I hope this is the right place to post this topic! I'm having some troubles with running Abinit in parallel, I'm now with v6.0.3 and I used --enable-mpi=yes --enable-mpi-io=yes --with-mpi-prefix=... during the configure step. Everything seems to work fine during a parallel job, but it looks like something is wrong with the output files.
The easiest way to test it is to use the tutorial "spin", more precisely the part "3. Another look at fcc Fe". When I run a serial job, I get exactly what is expected in the tutorial. When I run a parallel job however, despite the logfile and outfile are quite similar to those of the serial run, the DOS files are wrong: in the spin down channel, I only get a small signal in the s channel (the DOS in this l=0 channel integrates to 0.19 instead of 0.58 for the spin up channel) and other channels are just empty.
Can you reproduce this error? Do you have any hint?

Thanks in advance,
Regards,

Antonin

PS: extension restrictions on attachments are a pain :D I can't attach my output files!
Antonin Bourgeois
Postdoc @ Laboratoire de Physique des Solides, Université Paris-sud Orsay

Locked