[QE-users] Problem with QE 6.8 and libxc (rvv10-scan)

[Fabrizio Ferrari]

A small correction, in the two input_dft options I mentioned before it is
better to put 'vv10', not 'rvv10', since the correct name for that single
nonlocal term in qe is the former, even if it should work anyway.

Cheers,
Fabrizio

On Tue, Dec 21, 2021 at 12:10 PM Fabrizio Ferrari <
ferrariruffino.fz at gmail.com> wrote:

> Dear Christoph,
> I am having some hard times in reproducing your errors. A few things:
>
> - I get the segmentation fault with Libxc-6.0.0 (libxc-master), never with
> v5.1.7. By the way, since v6.0.0 is actually the develop branch, it changes
> from day to day (or week to week), so, if the library you are using has not
> been downloaded today, it may behave differently from the current one on
> gitlab. However, the error might be due to a change in the external
> parameters indexes with respect to v5.1.7. If so, you can try by commenting
> out the lines 1278,1280,1281,1282 in file XClib/dft_mod.f90 and by
> recompiling QE, but no guarantees.
>
> - With libxc-5.1.7 I get no errors with ifort. The rvv10-scan has the
> issue you mentioned, but the two alternatives work to me:
>   1) input_dft='mgga_x_scan mgga_c_scan rvv10'
>   2) input_dft='xc-000i-000i-000i-000i-263L-267L-rvv10'
>   If you have followed the compilation instructions in the user_guide of
> qe I don't think the problem is related to compilation.
>
> -in the new QE release wich will be available in a few days, the
> rvv10-scan option works as well, but among the two above input_dft
> alternatives only the second will be allowed.
>
> I hope this helps,
> Fabrizio
>
>
> On Tue, Dec 21, 2021 at 5:53 AM Christoph Wolf <wolf.christoph at qns.science>
> wrote:
>
>> Dear Fabrizio,
>>
>> Thank you for your reply and help. I tried to compile QE 6.8 with libxc
>> 5.1.7 but this leads to a core dump during execution with
>> 'rvv10-scan'/'MGGA_X_SCAN MGGA_C_SCAN'
>>
>>      Program PWSCF v.6.8 starts on 21Dec2021 at 13:49: 8
>>       ...
>>      Max number of k-points (npk) =  40000
>>      Max angular momentum in pseudopotentials (lmaxx) =  4
>>      *Segmentation fault (core dumped)*
>>
>> In a separate folder I compiled using the same procedure with libxc 6.0.0
>> and the core dump does not happen (but it crashes with the error in my
>> initial message).
>>
>>      Estimated max dynamical RAM per process >       3.08 GB
>>    Using LIBXC version       =    6   0   0
>> *pw.x: func_info.c:71: xc_func_info_get_ext_params_default_value:
>> Assertion `number >=0 && number < info->ext_params.n' failed.*
>>
>> In both cases I compiled libxc and QE with the new intel OneAPI (ifort
>> version 2021.5.0) and the "manual procedure" (by modifying the make.inc,
>> the --with-libxc --with-libxc-root etc does not detect 5.1.7 but it does
>> detect 6.0.0).
>>
>> Compiling libxc with intel and running the tests does not throw me any
>> error, so I wonder if I am doing something wrong when compiling QE... Maybe
>> some sneaky compiler option I have to consider?
>>
>> I did also try the *TPSS* meta-GGA and that works without any problem
>> (but I haven't checked if the output is meaningful, at least it did not
>> crash), so it could be a problem specific to the rvv10-scan...
>>
>> Best,
>> Chris
>> --
>> Group Leader "Theory of Quantum Systems at Surfaces"
>> IBS Center for Quantum Nanoscience
>> Seoul, South Korea
>>
>> _______________________________________________
>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
>> users mailing list users at lists.quantum-espresso.org
>> https://lists.quantum-espresso.org/mailman/listinfo/users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20211221/ae74cc66/attachment.html>