[Wien] WIEN2k Low CPU usage

[EB Lombardi]

Hello

The point of a scratch directory that is NOT NFS mounted is that large 
.vector files need not be copied back and forth on the NFS. So the 
scratch directory is any directory that exists separately on each node 
and is not NFS mounted, such as /tmp. This must have the same name on 
each node.

To use this, specify in your .bashrc which scratch directory Wien2k 
should  use:
Change the line
export SCRATCH=./                              # .vector files are 
placed in the case directory
to
export SCRATCH=/tmp  (or whatever)   # .vector files are placed in the 
/tmp directory.

(Log off and log back on for changes to take effect)

Make sure your scratch area has enough space for the .vector files.  It 
also needs to be cleaned up from time to time on each node.

Atta-fynn, Raymond wrote:
> Dear Sir,
>  
> I went through the WIEN mailing list and realized that
> you posted a problem concerning low cpu usage by lapw2. I
> posted a similar mail yesterday and got a partial answer from
> P. Blaha (suggesting that I create a $SCRATCH directory on 
> all nodes that a parallel job is running) but it is still not very clear 
> so I decided to contact you.
>  
> My question is how do I define a $SCRATCH  directory in
> each local node that I'm running a job and direct the
> case.vector and case.helpfiles there.
> Also, how do I tell lapw2 to read the vector files from the $SCRATH
> directory on each node?

Wien2k will do this automatically once the SCRATCH variable has been 
specified

>  
> Any help would be much appreciated. Thank you very much in advance.
>  
> R. Atta-Fynn
>  
>  
> ===================================================================
> Dr. Raymond Atta-Fynn
> Postdoctoral Research Fellow
> Deptartment of Physics Rm. 107 E
> The University of Texas at Arlington
> Arlington, TX 76019 
> Phone: 817-272-0621
PS: A Gigabit ethernet (or even a 100Mb network!) should be fast enough 
if the .vector files are not too big. In particular, the fact that you 
mention that lapw1 runs normally (which writes the .vector files), while 
lapw2 (which reads the same files) takes long, indicates that the 
problem may lie elsewhere?

I responded on the Wien2k Mailing list since other users may have 
similar problems.

Regards

Enrico Lombardi

-- 
Dr. E.B. Lombardi
Physics Department
University of South Africa
P.O. Box 392
0003 UNISA
South Africa
Tel: +27 (0)12 429-8654 / 8027
Fax: +27 (0)12 429-3643
e-mail: lombaeb at science.unisa.ac.za