[gpaw-users] vdw that memory intensive?

[Jens Jørgen Mortensen]

On 20-12-2011 15:56, Sampsa Riikonen wrote:
> Dear List,
>
> I have the following problem when trying to calculate a big water/ice
> clusters using the vdw functional.
>
> When trying to use the vdw module (commented lines correspond to older
> version of gpaw) ..
> ---------
> # from gpaw.vdw import FFTVDWFunctional #  gpaw<  0.8
> from gpaw.xc.vdw import FFTVDWFunctional # gpaw 0.8
> # vdw = FFTVDWFunctional(nspins=1, verbose=True) # gpaw<  0.8
> vdw = FFTVDWFunctional() # gpaw 0.8
> calc.set(xc=vdw)
> --------
> .. with a 20 water molecule cluster running on 16 prosessors each one of
> them having 4.5 gb of
> memory reserved, gpaw dies because it goes out of memory..! (Cray
> XT4/XT5, computer node linux: "OOM killer").

The default value for Nalpha is 20.

   
https://wiki.fysik.dtu.dk/gpaw/documentation/xc/vdw.html#perturbative-vdw-df-calculation-non-self-consistent

I think you can save some memory by matching this number to the number 
of processors.  So, you could try with 20 processors or use 16 
processors and reduce Nalpha to 16.

Jens Jørgen

> Has anyone experienced a similar problem..?
>
> _Without vdw_ that system runs ok.
>
> _With vdw_ for a single water molecule it also works ok.
>
> Regards,
>
> Sampsa
>
>
>
>
>
>
> _______________________________________________
> gpaw-users mailing list
> gpaw-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users