[Q-e-developers] PW vc-relax - FFT grids in final scf step

[Nicolas Mounet]

apologies, I just realized this fix was already done in the svn/git 
repository of the code.

cheers,
Nicolas

On 02/08/2017 01:20 AM, Nicolas Mounet wrote:
> Dear QE developers,
>
> I discovered recently that in QE 6.0 (pw.x), the FFT grids do not seem
> to be re-calculated and updated in the last final scf step of a vc-relax
> calculation done. Since it was not the case in the past, and I did not
> find any mention of it in the release notes, I presume this was not
> intended.
>
> The reason seems to be that in the routine PW/src/move_ions.f90, the
> nr[123] of the dfftp and dffts are not re-initialized before calling the
> function fft_type_allocate, which subsequently prevents the function
> realspace_grid_init to recalculate the grid.
>
> So, just adding the line
>
> dfftp%nr1=0; dfftp%nr2=0; dfftp%nr3=0; dffts%nr1=0; dffts%nr2=0;
> dffts%nr3=0
>
> right before the lines 335-336 in the PW/src/move_ions.f90 file
>
> CALL fft_type_allocate (dfftp, at, bg, gcutm, intra_bgrp_comm )
> CALL fft_type_allocate (dffts, at, bg, gcutms, intra_bgrp_comm)
>
> solves the problem.
>
> I attached the fixed function, plus an example with Si (with and without
> the fix - I used relaxed parameters and the PS library 1.0.0 PBE-US pseudo)
>
> Apologies in advance if this is not the proper way to submit a bugfix to
> the code, or if anything I wrote turns out to be wrong...
>
> Best regards,
> Nicolas Mounet
>
> --
> Nicolas Mounet
> EPFL STI IMX THEOS
> Station 9
> CH-1015 Lausanne (Switzerland)
> Phone: +41 21 69 31159
>
>
> This body part will be downloaded on demand.
>

-- 
Nicolas Mounet
Research Scientist
EPFL STI IMX THEOS
ME D2 1025
Station 9
CH-1015 Lausanne (Switzerland)
Phone: +41 21 69 31159