[Q-e-developers] Problem with trunk and intel-MPI

[Paolo Giannozzi]

I cannot reproduce this problem

On Fri, Apr 28, 2017 at 7:14 PM, Samuel Poncé <samuel.pon at gmail.com> wrote:
> Hello,
>
> I think any nscf calculation (after a scf to get density) should lead to
> this. Any call to read_file() so also post processing etc.
>
> We tested:
> 1) ifort (IFORT) 17.0.3 20170404 + Intel(R) MPI Library for Linux* OS,
> Version 2017 Update 2 Build 20170125
>
> 2) ifort (IFORT) 16.0.0 20150815 + Intel(R) MPI Library for Linux* OS,
> Version 5.1.1 Build 20150715 (build id: 12269)
>
> 3) ifort (IFORT) 16.0.0 20150815+  mvapich2 v 3.1.4
>
> What work:
>
> 1) GCC 4.9.2 + openmpi 1.8.4
>
> 2) intel 2017 and 2016 + openmpi 2.0.2
>
> Best,
>
> Samuel
>
>
> On 28 April 2017 at 18:04, Paolo Giannozzi <paolo.giannozzi at uniud.it> wrote:
>>
>> For any input? on specific inputs only? any compiler/mpi libraries?
>> some specific combination thereof?
>>
>> Paolo
>>
>> On Fri, Apr 28, 2017 at 6:39 PM, Martin Schlipf
>> <martin.schlipf at gmail.com> wrote:
>> > I tracked the issue and it happens somewhere in the subroutine
>> > read_file(). The error message that I get is
>> >
>> > PMPI_Bcast(1635): MPI_Bcast(buf=(nil), count=1, MPI_INTEGER, root=0,
>> > comm=0xc4000001) failed
>> >
>> > which looks to me like we try to broadcast an array that is not
>> > allocated.
>> >
>> > -Martin
>> >
>> > On 28/04/17 17:24, Paolo Giannozzi wrote:
>> >> I haven't noticed any such problem, but I haven't done extensive tests
>> >> recently. Without more information on what exactly fails how under
>> >> which circumstances, it is hard to say more
>> >>
>> >> Paolo
>> >>
>> >> On Fri, Apr 28, 2017 at 3:57 PM, Samuel Poncé <samuel.pon at gmail.com>
>> >> wrote:
>> >>> Dear developers,
>> >>>
>> >>> Martin and I are currently performing scaling tests with different
>> >>> compilers
>> >>> and libraries.
>> >>>
>> >>> We noticed that the current svn trunk work for scf calculation but not
>> >>> when
>> >>> restart (like nscf calculations) when using intel-MPI (we tested both
>> >>> 2016
>> >>> and 2017).
>> >>>
>> >>> The problem seems to also appear with mvapich2.
>> >>>
>> >>> This seems correlated with the recent introduction of default xml ?
>> >>>
>> >>> Note that the r13459 and before are fine with all of those.
>> >>>
>> >>>
>> >>> Best,
>> >>> Samuel
>> >>>
>> >>> --
>> >>>
>> >>>
>> >>> ------------------------------------------------------------------------------------------------
>> >>>      Dr. Samuel Poncé
>> >>>      Department of Materials
>> >>>      University of Oxford
>> >>>      Parks Road
>> >>>      Oxford OX1 3PH, UK
>> >>>
>> >>>      Phone: +44 1865 612789
>> >>>      email: samuel.ponce at materials.ox.ac.uk
>> >>>      web:
>> >>> http://giustino.materials.ox.ac.uk/index.php/Site/SamuelPonc%e9
>> >>>
>> >>> ------------------------------------------------------------------------------------------------
>> >>>
>> >>>
>> >>> _______________________________________________
>> >>> Q-e-developers mailing list
>> >>> Q-e-developers at qe-forge.org
>> >>> http://qe-forge.org/mailman/listinfo/q-e-developers
>> >>>
>> >>
>> >>
>> >
>> > _______________________________________________
>> > Q-e-developers mailing list
>> > Q-e-developers at qe-forge.org
>> > http://qe-forge.org/mailman/listinfo/q-e-developers
>>
>>
>>
>> --
>> 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
>>
>> _______________________________________________
>> Q-e-developers mailing list
>> Q-e-developers at qe-forge.org
>> http://qe-forge.org/mailman/listinfo/q-e-developers
>
>
>
>
> --
>
> ------------------------------------------------------------------------------------------------
>     Dr. Samuel Poncé
>     Department of Materials
>     University of Oxford
>     Parks Road
>     Oxford OX1 3PH, UK
>
>     Phone: +44 1865 612789
>     email: samuel.ponce at materials.ox.ac.uk
>     web: http://giustino.materials.ox.ac.uk/index.php/Site/SamuelPonc%e9
> ------------------------------------------------------------------------------------------------
>
>
> _______________________________________________
> Q-e-developers mailing list
> Q-e-developers at qe-forge.org
> http://qe-forge.org/mailman/listinfo/q-e-developers
>



-- 
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