[Wien] parallel calc. without NFS

[Kevin Jorissen]

Hi Stefaan, wake up ! ;-)  We've had this conversation before.
If the number of parallel jobs is not equal to the number of machines, then
we have to make sure lapw2 uses exactly the same machine for each particular
job as lapw1, otherwise we may get crazy results.  So :
*either you have to rewrite lapw2para to take care of this (= what I did)
*or you have to make sure that number_of_jobs=number_of_machines (=what you
do), by setting the weights for the machines in such a way that the sum of
all weights is equal to the number of k-points.


Kevin.

----- Original Message -----
From: "Stefaan Cottenier" <Stefaan.Cottenier at fys.kuleuven.ac.be>
To: <wien at zeus.theochem.tuwien.ac.at>
Sent: Wednesday, July 09, 2003 10:43 AM
Subject: Re: [Wien] parallel calc. without NFS


>
>
> > I started to use local SCRATCH directories on our machines some months
ago
> > and may be able to help you if you run into problems there (some scripts
> > need to be rewritten).
>
> Hello Kevin,
>
> I use local scratch with the unmodified code (apart from some extra
delay).
> What did you have to adapt in which scripts? (maybe parts that do not
belong
> to the regular scf-cycle ?)
>
> Stefaan
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>