Page 1 of 1

Issue with parallelization when calculating dielectric matrix

Posted: Tue Mar 04, 2025 8:41 pm
by sophie_weber

Hello,

I am trying to calculate a frequency-dependent dielectric function with LOPTICS=True , and if I try to parallelize by setting NCORE or NPAR I get the following error:
VASP internal routines have requested a change of the k-point set. Unfortunately, this is only possible if NPAR=number of nodes. Please remove the tag NPAR from the INCAR file and restart the calculation.

I see in earlier posts that this is a bug for DFPT routines for some versions of VASP. I am using VASP version 6.3, can you please verify whether this error has been fixed in more recent versions?

I need to calculate off-diagonal elements and hence include local field effects, to the calculations are really too expensive to run without better band parallelization, or at least with the settings I have (including a sufficiently large number of virtual bands for LOPTICS=true) the job keeps crashing from memory issues if I don't set NCORE.

Thank you in advance!


Re: Issue with parallelization when calculating dielectric matrix

Posted: Fri Mar 07, 2025 8:12 am
by ferenc_karsai

You wrote that you ran LOPTICS=.TRUE. calculations. They work fine with NCORE>1.

I guess you meant calculations with LEPSILON=.TRUE.. For those calculations I get what you reported. So please don't set NCORE or NPAR in the INCAR file when you run these calculations.

I'm not sure if this is a bug or NCORE>1 is just simply not implemented (like for LPEAD and LCALCEPS). I will ask my colleagues about this.


Re: Issue with parallelization when calculating dielectric matrix

Posted: Fri Mar 07, 2025 8:35 am
by sophie_weber

Ah, thanks a lot, I didn't even check with LOPTICS because I assumed it would be the same issue.