[Wien] QTL after parallel SCF

[Robert Laskowski]

hi,
lapw2 in qtl mode run as a single program, therefore it needs vectores stored locally. 
You can  set  SCRATCH in your terminal to ./ and rerun lapw1, or just cp the vectores from remote nodes to 
local node. The just run lapw2 -qtl .... Look at the uplapw2.def not upqtl.def 

regards

Robert   

On Monday 16 June 2008 10:42:22 Dan Maftei wrote:
> Dear all,
> 
> I am performing a spin-polarized SCF run in parallel over k-points (20 
> jobs on 10 dual-CPU nodes). The cluster nodes share the working 
> directory via NFS, the $SCRATCH directory being local (under /tmp). 
> 
> lapw1 (-up/-dn)  generates 20 case.vector_* files (2 on each node) for 
> spin up/dn in the corresponding SCRATCH dirs. So far, everything works 
> fine.
> 
> The problem arises when trying to compute partial charges needed for the 
> DOS: when running "x lapw2 -qtl -up -p".
> 
> lapw2 complains about the vector files and I've noticed in "upqtl.def" 
> that it refers to case.vectorup and case.vectordn instead of (I guess) 
> case.vectorup_* and case.vectordn_*.
> 
> Should I combine the case.vectorup_*/case.vectordn_* stored on $SCRATCH 
> dir of the nodes into 2  coresponding vector files tor each spin, in 
> order to compute the DOS? If so, how to do that?
> 
> Regards,
> 
> Dan Maftei.
> Assistant Professor,
> Department. of Chemistry,
> "Al. I. Cuza" University,
> Iasi-Romania.
> 



-- 
Dr Robert Laskowski

Vienna University of Technology, 
Institute of Materials Chemistry, 
Getreidemarkt 9/165-TC, A-1060 Vienna

tel. +43 1 58801 15686
Fax  +43 1 58801 15698