[Q-e-developers] problem with pressure calculation -- cp.x (bug?)

[Filippo SPIGA]

Hello Pier Luigi ,

I forward your email to the q-e-developers mailing-list. I am not expert of the physics, I am more on the "computing/HPC" side of things. Hopefully someone can clarify this point of you. If it is a bug (it looks like a bug and it smells like a bug.... so probably we need to investigate properly) we will work to fix it by the official release My uninformed guess is that the quantities that varies across different machines is calculated heavily using some libraries or some OpenMP. This is something easy to check. I cannot comment about cp.x versus pw.x correctness.

Anyone has a guess about these discrepancies or a practical suggestion how to verify this?


On Sep 13, 2016, at 7:21 AM, Pier Luigi Silvestrelli <pierluigi.silvestrelli at pd.infn.it> wrote:

I have found that in QE (in the new 6.0-beta version
and in previous versions as well) the pressure calculation seems
to be incorrect using 'cp.x' in combination with non-local vdW-corrected
functionals, such as vdW-DF, vdW-DF2, vdW-DF-cx, rVV10,...,
as verified by comparing results obtained using both 'pw.x' and
'cp.x' programs.
For instance, considering a single liquid-water configuration
and the vdW-DF functional, 'pw.x' (both using the 'scf' and 'md'
options) gives P=-3.55 kbar, while 'cp.x' gives instead P=200.99 kbar.
Moreover, P computed by 'cp.x' using the same version of the code
on different machines gives different results, while all the other
quantities (total energy,...) are identical.

I would like to let you know this problem, which is of some importance
for efficient vdW-corrected ab initio MD simulations of liquids.

Thanks in advance. Ciao.

Pier Luigi Silvestrelli

Dipartimento di Fisica e Astronomia G. Galilei
Via Marzolo 8
I-35131 Padova, Italy
tel. +39-049-8277171
fax  +39-049-8277102
e-mail : psil at pd.infn.it

--
Filippo SPIGA ~ Quantum ESPRESSO Foundation ~ http://www.quantum-espresso.org