[Wien] iter. diag and scratch files

Peter Blaha pblaha at theochem.tuwien.ac.at
Thu Dec 13 20:14:00 CET 2012


I don't know how many cycles you get in 48 h.

Usually one can use iter.diag without restarting it for quite some time.
I usually restart it (i.e. making a full diagonalization) only once in a normal
scf-cycle (usually after the first big changes are over), an in a structural
optimization also only once or twice, just to make sure that I do not have a
non-converging case (which can happen for special cases) with iter.diag.

If you do not encounter problems, I'd not restart, i.e. I'd not delete the old vectors
(scratch-directory).

Check your timing on how much faster the later scf cycles are compared to the
full diag.
You are  always wasting this time when restarting without need.

Am 13.12.2012 16:05, schrieb Francisco Garcia:
> Dear users,
>
> I am running iterative diagonalization calculations on very big slab
> at a shared user computational facility. At the end of each run (the
> max runtime is 48 hours) , I delete the scratch directory (whether or
> not the job is converged). When I re-start the run, I create a new
> scratch directory.
>
> My question is the following: When running iterative diag.
> calculations, should I keep the initial scratch directory (the scratch
> directory created during the first run) intact and use it for
> subsequent runs or it is okay to delete and make a new directory per
> run as I have described above.
>
> Thank you.
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>

-- 
-----------------------------------------
Peter Blaha
Inst. Materials Chemistry, TU Vienna
Getreidemarkt 9, A-1060 Vienna, Austria
Tel: +43-1-5880115671
Fax: +43-1-5880115698
email: pblaha at theochem.tuwien.ac.at
-----------------------------------------


More information about the Wien mailing list