What you wrote is right...and wrong.

If you are just running 1 process at a time in the non-parallel mode, what
you wrote is correct.

However, in many cases one either runs  many different processes
simultaneously, often as many as you have cores (ignoring hyperthreading
which is rarely useful in my experience). For larger calculations you may
also want to use mpi in which case every core will have the same memory

You have a relatively small cluster, so I suggest that you leave it with
the default values, particularly if you are relatively new to Wien2k. While
the values you used probably won't matter, they are not that safe. I
suggest leaving them for now and running a number of test calculations.

On Wed, Nov 4, 2015 at 8:04 AM, tongwei <weitong...@hotmail.com> wrote:

> Thanks.
> For the NMATMAX and NUME values, I am following the prompt during
> installation:
> "NMATMAX=10000 ==>   1GB (real) (==> cells with about 50-150
> atoms/unitcell)
>     ==> for lapw1c:    NMATMAX will be reduced internally to NMATMAX/sqrt2
>     ==> for lapw1_mpi: NMATMAX will be increased internally to
> NMATMAX*sqrt(NP)
> NUME determines the number of states to output. As a rule of thumb one can
> estimate 100 basis functions per atom in the cell and 10 occupied states
> per
> atom, so set    NUME=NMATMAX/10.
> "
> I have 128GB at node1 and 64GB memory at node2 . So, I assume that 32GB
> should be available for WIEN2k. Then 32*10000=320000, NUME=NMATMAX/10=32000.
> Is this wrong? Then, where to change this value?
> Best,
> WT
> ------------------------------
> Date: Mon, 2 Nov 2015 09:01:41 -0600
> From: l-ma...@northwestern.edu
> To: wien@zeus.theochem.tuwien.ac.at
> Subject: Re: [Wien] Some error messages during Installation of WIEN2k_14.2
> on CentOS6.5
> I am 99.9% certain that none of these is an error. The compile script will
> tell you if there are any at the end, these are flagged with "Error".
> N.B., the values
> set value for NMATMAX=320000
> set value for NUME=32000
> are large, you could run into problems with these if you are running
> k-point parallel for larger jobs. Since you only have a 2 node cluster with
> 1G ethernet (?) you may be only intending to run smaller jobs in which case
> it will be fine. You could get an OOM (Out Of Memory) OS crash with larger
> jobs.
> --
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
Corrosion in 4D: MURI4D.numis.northwestern.edu
Co-Editor, Acta Cryst A
"Research is to see what everybody else has seen, and to think what nobody
else has thought"
Albert Szent-Gyorgi
Wien mailing list

Reply via email to