[Pw_forum] Compilation with intel compiler and MKL.

joaquin peralta jperaltac at gmail.com
Tue Feb 2 18:35:39 CET 2016


Dear Paolo,

Setting the attr='' doesn't work, however comment out the call works like a
charm. So far it is probably a compiler/library/OS issue, because we tried
in a Debian 7.9 machine (with the same intel and intelmpi), and works
perfectly. The machine with the problem is a Debian testing.

Thanks in advance for all your help,
Joaquin

On Tue, Feb 2, 2016 at 2:15 PM Paolo Giannozzi <p.giannozzi at gmail.com>
wrote:

> Since you are not reading from xml, try to set attr=' ' before that call,
> or to comment out the call. The only reason I can think of (that doesn't
> imply a compiler/library/mpi weirdness) is that attr may not be set to
> anything if reading from a text file, and sometimes code get confused if a
> non-initialized variable is broadcast
>
> Paolo
>
> On Tue, Feb 2, 2016 at 6:07 PM, joaquin peralta <jperaltac at gmail.com>
> wrote:
>
>> Dear Paolo,
>>
>> Thank you very much for your promptly response. I found the line where
>> the code freezes.
>>
>> CALL mp_bcast( attr, ionode_id, intra_image_comm )
>>
>> However I can't figured out what could be the reason, or how to avoid the
>> hung.
>>
>> Do you have any additional suggestion?
>>
>> Thanks in advance,
>> Joaquin Peralta
>> Assitant Professor,
>> Faculty of Physics,
>> Exact Science Department,
>> Universidad Andres Bello,
>> Santaigo, Chile.
>>
>>
>> On Tue, Feb 2, 2016 at 1:23 PM Paolo Giannozzi <p.giannozzi at gmail.com>
>> wrote:
>>
>>> Go to Modules/read_input.f90, where routine "open_input_file" is called
>>> (should be line 50), and follow what happens later and where the code
>>> freezes. Just add some prints: there are smarter ways to figure out where a
>>> code is hung, but I don't remember how to do it. Once you find the exact
>>> line where it freezes, either you easily find why, or, if there is no
>>> obvious reason, it is likely a problem of your compiler/ libraries  / mpi.
>>>
>>> Paolo
>>>
>>>
>>> On Tue, Feb 2, 2016 at 4:40 PM, joaquin peralta <jperaltac at gmail.com>
>>> wrote:
>>>
>>>> Dear Youssef,
>>>>
>>>> I did the compilation as you indicated, however my QE-run frozen with
>>>> more than one CPU. I have the intel parallel studio 2016 and the impi 5.1.
>>>> After compile I did :
>>>>
>>>> mpirun -np 1 /path/to/pw.x -input scf.in     <-This run perfectly
>>>>
>>>> However :
>>>>
>>>> mpirun -np 2 /path/to/pw,x -input scf.in      <- Doesn't work!!!, It
>>>> frozen in "Reading input from scf.in"
>>>>
>>>> Do you have any suggestion or hint about what could be the reason
>>>> because is not working?
>>>>
>>>> I have all the exports from intel-suite in my bashrc :
>>>> source /opt/intel/bin/compilervars.sh intel64
>>>> source /opt/intel/mkl/bin/mklvars.sh intel64
>>>> source /opt/intel/impi/5.1.2.150/bin64/mpivars.sh
>>>>
>>>> And I also set my stack size to unlimited in the kernel settings.
>>>>
>>>> Is there a code issue? Or is the intel mpi?
>>>>
>>>> Thanks in advance,
>>>> Joaquin Peralta
>>>> Assitant Professor,
>>>> Faculty of Physics,
>>>> Exact Science Department,
>>>> Universidad Andres Bello,
>>>> Santaigo, Chile.
>>>>
>>>>
>>>> On Tue, Feb 2, 2016 at 9:57 AM Youssef <aharbil at gmail.com> wrote:
>>>>
>>>>> Dear
>>>>>
>>>>>
>>>>>
>>>>> *Yeonung, I used to compile qe-5.30 with intel parallel studio 2016,
>>>>> using this statements: *
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> *export FC=mpiifort export F90=mpiifort export F77=mpiifort export
>>>>> MPIF90=mpiifort export CC=mpiicc export CPP="icc -E" export CFLAGS=$FCFLAGS
>>>>> export AR=xiar export BLAS_LIBS="" export
>>>>> LAPACK_LIBS="-lmkl_blacs_intelmpi_lp64" export
>>>>> SCALAPACK_LIBS="-lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64" export
>>>>> FFT_LIBS="-L$MKLROOT/intel64" ./configure  --enable-parallel *
>>>>> you can save it as a bash script, an run it inside qe folder.
>>>>>
>>>>> Of course  make sure that the intel parallel studio wrapper is in your
>>>>> path (e.g. :  source /opt/intel/bin/compilervars.sh intel64 ), this  invoke
>>>>> all the needed environmental variables.
>>>>>
>>>>> Youssef Aharbil,
>>>>>
>>>>> Laboratory of Physics and Chemistry of Materials
>>>>>
>>>>> Faculty of sciences Ben msik, Casablanca
>>>>>
>>>>> Morocco
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> 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
>>>
>>> _______________________________________________
>>> 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
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20160202/52b04e52/attachment.html>


More information about the users mailing list