[Pw_forum] ESM error in qe 6.0

[Paolo Giannozzi]

Hi Ale, I think all versions after luly 2015 have the same problem. It
is related to OpenMP: the "normal" one-dimensional FFT routine,
cft_1z, has an internal OpenMP loop, while in ESM there are OpenMP
loops calling one-dimensional FFT's. The workaround is to have a copy
of cft_1z (esm_cft_1z) that doesn't do OpenMP and a thread-safe FFT
initialization (esm_cft_1z_init). These routines have fallen out of
sync with the rest of the code and do not work for recent FFT
extensions (notably, DFTI routines from Intel MKL, likely the one that
Cineca uses).

I suspect that in the absence of OpenMP parallelization, the code
works if you replace esm_cft_1z with cft_1z and comment out
esm_cft_1z_init, but I haven't tried it. Better solutions are
relatively simple but require a preliminary and far-reaching decision
on how to perform mixed MPI+OpenMP parallelization for FFT's.

Paolo

On Sat, Dec 17, 2016 at 1:38 PM, Alessandra Satta
<alessandra.satta2015 at gmail.com> wrote:
> Dear Dr. Otani,
>  thank you so much for your prompt reply.
>  I will be waiting for.
>  Let me ask you one more question: does this mean that none of the previous
> QE-releases work with this card?
>  May I try with the 5.4?
>  Thanks again,
>   alessandra
>
>
> Alessandra Satta
> CNR-IOM UOS Cagliari
> c\o Dipartimento di Fisica dell’Universita' di Cagliari
> Cittadella Universitaria - SP Monserrato-Sestu Km. 0,700
> I-09042 - Monserrato (CA)
> Tel.: +39 070 675 4865
> Fax : +39 070 675 4892
>
>
>
> On 17 Dec 2016, at 13:29, Minoru Otani <minoru.otani at aist.go.jp> wrote:
>
> Dear Dr. Alessandra Satta,
>
> Now we are working on this problem. Please wait a few more weeks.
> The revised version of the esm.f90 will be sent to Paolo.
>
> Best regards,
> Minoru
>
> --------------------------------------------------------------------------------------------------------
> Minoru Otani
> National Institute of Advanced Industrial Science and Technology
> Research Centre for Computational Design of Advanced Functional Materials
> email : minoru.otani at aist.go.jp
> tel : +81-29-861-5202
> fax : +81-29-861-3171
> ---------------------------------------------------------------------------------------------------------
>
> On Dec 17, 2016, at 8:48 PM, Alessandra Satta
> <alessandra.satta2015 at gmail.com> wrote:
>
> Dear QE-users,
> I’m currently having the same error message mentioned below (Error in
> routine esm_cft_1z_init (1): no scalar fft driver specified) in the qe 6.0
> version. Same kind of calculation (ESM).
> Trying to follow Paolo’s suggestion but unable to have a successful
> compilation on cineca’s MARCONI (or GALIEO). My fault I guess, I’m not
> loading the right modules.
> Is there a (week-end worker).and.(CINECA user)  having a make.inc suited for
> CINECA’s libraries that includes scalar FFTs.
> I would very much appreciate it,. Thank you,
>   alessandra
>
> Alessandra Satta
> CNR-IOM UOS Cagliari
> c\o Dipartimento di Fisica dell’Universita' di Cagliari
> Cittadella Universitaria - SP Monserrato-Sestu Km. 0,700
> I-09042 - Monserrato (CA)
> Tel.: +39 070 675 4865
> Fax : +39 070 675 4892
>
>
> On 17 Sep 2016, at 08:28, Paolo Giannozzi <p.giannozzi at gmail.com> wrote:
>
> Currently you cannot use ESM together with FFT in the Intel MK library
> (DFTI). Use FFTW or FFTW3 instead
>
> Paolo
>
> On Fri, Sep 16, 2016 at 10:08 PM, Nkosinathi Malaza <nuttymalaza at gmail.com>
> wrote:
> Dear All,
>
> I am trying to run calculations using the ESM mode, the error message that I
> get in the output file is:
>
> Error in routine esm_cft_1z_init (1): no scalar fft driver specified
>
> How can I resolve this?
>
> Thank you
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
> --
> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> Phone +39-0432-558216, fax +39-0432-558222
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
>
>
>
>
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum



-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222