[Q-e-developers] Compilation of QE with PGI 15

[Paolo Giannozzi]

Maybe this (from install/m4/x_ac_qe_mpif90.m4) is no longer true?

# check serial Fortran 90 compiler. This must be done before performing
# the check for the parallel compiler (section below) because option
# --disable-parallel will do very strange things otherwise. The reason
# seems to be that autoconf does not repeat all tests for the second
# occurrence of AC_PROG_FC. So the first occurrence is the one that
# must always be performed, the second is optional. PG & CC sep.2006

# use F90 if set
if test "$f90" = "" ; then f90="$try_f90" ; fi
AC_PROG_FC($f90)
f90=$FC
AC_FC_SRCEXT(f90)

There is a lot of obsolete stuff in configure that should be removed.

Paolo

On Mon, Jun 6, 2016 at 7:05 PM, Martin Schlipf <martin.schlipf at gmail.com> wrote:
> Investigating a bit more, I think the issue is related to autoconf.
> Since version 2.64, autoconf expands commands before the file that
> requires them
> http://www.gnu.org/software/autoconf/manual/autoconf.html#Expanded-Before-Required
>
> In this particular case, AC_PROG_FC is expanded before the rest of
> X_AC_QE_MPIF90. As a consequence there is a large additional part in the
> code directly after
> # Checking MPIF90...
> but before the search for the compilers with try_f90.
>
> The problem is that the additional parts include a different search for
> a compiler with a different order. Instead of ifort, pgf90, nagfor, ...,
> gfortran, it checks for gfortran first. Hence, it doesn't come to the
> point where it would check for pgf90.
>
> I don't know how to fix the issue in autoconf, but one can delete the
> additional lines in the resulting configure file and then the PGI
> compiler works.
>
> -Martin
>
> On 06/06/16 13:43, Paolo Giannozzi wrote:
>> The configure file currently on svn was updated in rev.12409. The only
>> changes to configure since rev.12409 are
>>    rev. 12416: removal of "-C" from cppflags (required by PGI compiler)
>>    rev. 12419: proposed by David Strubbe: see
>> https://www.mail-archive.com/pw_forum%40pwscf.org/msg28594.html
>> I don't know if the latter is responsible for the change of behavior
>> you observe. Please try "svn update -r 12416 configure.ac
>> m4/x_ac_qe_f90.m4   m4/x_ac_qe_mpif90", refresh configure, report.
>> What is "configure.new" good for, by the way?
>>
>> Paolo
>>
>> On Mon, Jun 6, 2016 at 12:29 PM, Martin Schlipf
>> <martin.schlipf at gmail.com> wrote:
>>> Dear Filippo,
>>>
>>> we thought for some time that it was working with the added
>>> refresh-configure step. But now we realized that the refresh-configure step
>>> leads to gfortran being used instead of the PGI compiler. We also found out
>>> that the reason the NAG compiler doesn't work is the same as for the PGI
>>> compiler, i.e., the cpp -P -C choice in the make.sys file.
>>>
>>> We don't understand why the PGI compiler is not chosen after
>>> refresh-configure, so we provide the changed configure and the reference
>>> configure file so that you may have a look.
>>>
>>> Best regards
>>> Martin & Samuel
>>>
>>>
>>> On 24/05/16 17:35, Filippo SPIGA wrote:
>>>> On May 24, 2016, at 4:39 PM, Samuel Poncé <samuel.pon at gmail.com> wrote:
>>>>> Can I run the script ./refresh-configure.sh and commit to the trunk?
>>>> Yeah, it is not a problem.
>>>>
>>>> My personal suggestion is that if you need a refresh configure for your
>>>> BuildBot then it is more flexible for you to add an additional step in the
>>>> BuildBot itself. Very trivial thing to do, I did it to generate snapshots:
>>>> http://qe-forge.org:8010/builders/QE-FORGE%20%28PACKAGING%29/builds/6
>>>>
>>>> --
>>>> Mr. Filippo SPIGA, M.Sc.
>>>> Quantum ESPRESSO Foundation
>>>> http://www.quantum-espresso.org ~ skype: filippo.spiga
>>>>
>>>> *****
>>>> Disclaimer: "Please note this message and any attachments are CONFIDENTIAL
>>>> and may be privileged or otherwise protected from disclosure. The contents
>>>> are not to be disclosed to anyone other than the addressee. Unauthorized
>>>> recipients are requested to preserve this confidentiality and to advise the
>>>> sender immediately of any error in transmission."
>>>>
>>>>
>>>> _______________________________________________
>>>> 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
>>>
>>
>>
>
> _______________________________________________
> 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