[Wien] insufficient virtual memory

Peter Blaha pblaha at theochem.tuwien.ac.at
Tue Jun 11 08:59:15 CEST 2013


You did not tell us which hardware you are using.

Eventually, you can do it if you make sure that you are running only
ONE k-point on this machine; i.e. do not use k-parallelism (at least not 
on the same computer).

PS: You want to simulate a single sheet of graphene. So why would you 
use more than ONE k-point in the z-direction (9x9x2) ????

On 06/11/2013 07:54 AM, Hajar Nejati wrote:
> dear Oleg
> I mention the points that I think can be benefit to solve my problem:
>
> 1. The origin cell has these lattice constant: a=b=2.53 and c=15.0 Angstrom.
> 2. 4×4×2 refer to the dimensions of supercell while the k-mesh is 24
> k-points in irreducible Brillouin zone (9×9×2).
> 3. I run lapw1 -nmat_only then I saw 14136 in case.nmat_only file (that
> is greater than 10000). So I had to increase NMATMAX to 20000 and
> compile wien2k again and rerun the scf calculation with lower RKMAX (=2).
>
> Is there any suggestion except considering a parallel MPI version? (I
> use .machines file to do parallel calculation)
> Thanks a lot, again.
> Best Regards
> Hajar
>
>
> ------------------------------------------------------------------------
> *From:* Oleg Rubel <orubel at lakeheadu.ca>
> *To:* wien at zeus.theochem.tuwien.ac.at
> *Sent:* Monday, June 10, 2013 1:17 PM
> *Subject:* Re: [Wien] insufficient virtual memory
>
> Dear Hajar,
>
>  > I was trying to do a supercell calculation (hydrogenated graphenewith 64
>  > atoms) within 4×4×2.
>
> It is unlikely to run such a job on a single processor. You need to
> consider a parallel MPI version (see Sec. 5.5 of the UG).
>
> Does 4×4×2 refer to the k-mesh? If yes, this is too dense. Most likely
> you can go with a single k-point.
>
>  > When I run scf, this message appears:
>  > Lapw0 end
>  > severe (41): insufficient virtual memory
>  > lapw1c unknown
>  > lapw1c unknown
>  > lapw1c unknown
>  > (To be sure, I have searched all the previous emails on this topic, but
>  > failed to solve my problem.)
>  > I know that RKMAX should be small and NMATMAX large (I increase NMATMAX
>  > from 10000 to 20000 and NUME to 1500). The scf runs without error
>  > massage when I decrease RKMAX to 2, but the total energy NOT converge
>  > (when I set RMT of H atoms to 0.6, the error remains).
>
> The purpose of NMATMAX is to prevent the calculation from crashing when
> the matrix size increases. Wien2k does it by lowering the effective
> RKmax, which you can find in the *scf file. However, changing NMATMAX is
> not a solution for this problem.
>
> I would also suggest to reduce the disparity between RMTs as much as you
> can.
>
> I hope this will help.
>
>
> Best regards
> Oleg
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at <mailto:Wien at zeus.theochem.tuwien.ac.at>
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:
> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>
>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300             FAX: +43-1-58801-165982
Email: blaha at theochem.tuwien.ac.at    WWW: 
http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------


More information about the Wien mailing list