[Wien] Parallelizing UP and DN for Fix-spin-moment calculation

[Laurence Marks]

I am 98% certain that you can change the order in runfsm to

lapw1cup ---> lapw1cdn ---> lapw2cup ---> lapw2cdn

I suggest that you create a small script to verify before running the
calculation in anger. However, be aware that runfsm may not get the
energy right, so is probably only useful to force a particular spin
state.

N.B. it's not too hard to edit lapw1para and lapw2para to run up/dn in
parallel using .machinesup and .machinesdn. I have unsafe versions
myself -- unsafe because they will probably die with -so or other
options but if anyone wants to clean them up let me know.

On 2/23/07, Chiung-Yuan Lin <cylinwn at gmail.com> wrote:
> Dear all,
>   Some time ago, I asked how to parallelize spin-up and dn jobs for a
> runsp_lapw calculation. And Prof. Blaha showed me a straightforward
> way to do it by modifying
> the runsp_lapw script:
>
> total_exec    lapw1 $it0 -up $para $nohns &
> ...
> total_exec    lapw1 $it0 -dn $para $nohns &
>
> Recently I am planning a runfsm_lapw calculation and also want to parallelize
> spin-up and dn jobs. However, the simple way applied to runsp_lapw doesn't
> seem to apply to runfsm_lapw since the job flow goes like
>
> lapw1cup ---> lapw2cup ---> lapw1cdn ---> lapw2cdn
>
> Does anyone have an idea how to parallelize spin-up and dn in runfsm_lapw?
> I am running a 4-kpoint case on a shared-memory machine with 8 processors.
> If I can parallelize spin-up and dn as well as the k points, the jobs
> fit into the
> machine perfectly.
>
> Thanks,
> Chiung-Yuan
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>


-- 
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60208, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
Web: www.numis.northwestern.edu
EMM2007 http://ns.crys.ras.ru/EMMM07/