The ABINIT Users Mailing List ( CLOSED )

[Tianjiao Zhang <tjzhang@quantumstates.com>]

Hi,
Just some tips.
(1) Make sure that your parallel environment is set up using lamboot. 
This can be checked by looking
back how you compile the parallel source code to get the abinip. I think 
if you get a mpi compiler there, the running tools should also be provided.
(2) If you do use lamboot,
see the user guide at
http://www.lam-mpi.org/using/docs/
for LAM/MPI User's Guide
It's worth tons of time later.
(3) It's difficult to know which OS you are using from your email. ( 
What is an IBM cluster ?)
 You can get it by type
uname -a    
....
Hopefully you can the node names from the file
/etc/hosts

(4) You may also need to setup rsh or ssh etc to get it work.

(5) Ask your vendor for a documentation about how to run parallel jobs.

(6) See your system administrator for help

(7) If you are using MPICH package for your MPI parallel environment, 
your can look at its manual
at
http://www-unix.mcs.anl.gov/mpi/
or just type
man mpirun  
( hope it installed)

Personally, I just run my parallel job as:

PATH_TO_MPIRUN/mpirun -arch p4 -np 6 -machinefile mymachinelist 
PATH_TO_abinip/abinip <inputfiles >& log

The mymachinelist  is just a text file containing each host name  or ip 
address per line . To tune it for better performance, read manuals.

(8) Know a little about the file system setup on your cluster.  Pay 
attention that NFS writing is slow after all.

(9) Good luck !


Tianjiao

Chun Li wrote:

> Dear abinit users,
>
> I am trying to use the parallel version of abinit on an IBM cluster; I
> followed the paral_use file in the info directory. When I ran the
> command lamboot cluster_file, I have no idea that how to get the file
> that contains the name of machines. Is it just written by oneself with
> fixed format (then what's the format)? Or there is ready such file in
> some directory? Thanks a lot.
>
> Best regards.
>
> Chun Li
>
>
>
>