[Q-e-developers] [Pw_forum]

[Samuel Poncé]

Dear Paolo,

Thank you for your reply. I'm actually trying to set up a test farm using
buildbot for EPW (current specs of the test farm available at
http://epw.org.uk/Main/TestFarm but should increase).
The buildbot waterfall is currently inside the oxford network only. We
should open it when everything compile. So far I, everything work except
intel 13.1.3 and pgi 12.

Since EPW rely on QE (EPW now work with the latest trunk of QE), it has the
side effect of testing QE as well. The plan is only to test the configure &
make of QE and EPW and to test physical results of EPW only.

I have relaunch the buildbot slaves (r-11677) and pgi is now happy with
your changes to generate_function.F90. Thank you.

Actually, pgcc still crashed later in

make[1]: Entering directory
`/home/buildbot2/farm2-slave1/EPW-farm2_pgi12/QE/clib'
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c
customize_signals.c
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c stack.c
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c c_mkdir.c
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c copy.c
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c cptimer.c
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c eval_infix.c
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c fft_stick.c
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c indici.c
pgcc -fast -D__PGI -D__FFTW -D__MPI -D__PARA  -I../include  -c
md5.cPGC-S-0040-Illegal use of symbol, __SWBLK_T_TYPE
(/home/farm2/local/pgi_1210/linux86-64/12.10/include/bits/types.h:
159)
PGC-W-0156-Type not specified, 'int' assumed
(/home/farm2/local/pgi_1210/linux86-64/12.10/include/bits/types.h:
159)
PGC-S-0040-Illegal use of symbol, __syscall_slong_t (/usr/include/time.h: 123)
PGC-W-0156-Type not specified, 'int' assumed (/usr/include/time.h: 123)

I haven’t yet figure out why.

The crash on intel 13 occurs in the routine ESM you are mentioning:

force_lc.f90(20): error #6580: Name in only-list does not exist.
[ESM_FORCE_LC]
  USE esm,       ONLY : esm_force_lc, do_comp_esm, esm_bc
------------------------^
compilation aborted for force_lc.f90 (code 1)


This crash is weird since the esm_force_lc is part of the esm module and is
public.

Best,

Samuel

On 14 August 2015 at 20:54, Paolo Giannozzi <p.giannozzi at gmail.com> wrote:

> Thank you Samuel. Guess what? some compilers may not like your change,
> because they consider erf and erfc as intrinsic. Or at least, there used to
> be such problem and this is why QE uses (almost) everywhere (see note
> below) functions qe_erf, qe_erfc, defined in flib/erf.f90 (maybe also for
> another reason: buggy intrinsic erf). Please try now (i just made such
> change).
>
> Anybody willing to try if all compilers now support the same syntax for
> intrinsic functions erf/erfc is tronlgu encouraged to do so.
>
> Note: in ESM, there is a different function "esm_erfc". Not sure whether
> this can be merged with/into qe_erfc, I'll inquire
>
> Paolo
>
> On Fri, Aug 14, 2015 at 8:16 PM, Samuel Poncé <samuel.pon at gmail.com>
> wrote:
>
>> Dear QE developers,
>>
>> I've been trying to compile the current trunk of QE using the pgi 12.10
>> compiler.
>>
>> The compiler was not happy about the following:
>> PGF90-S-0038-Symbol, erfc, has not been explicitly declared
>> (generate_function.F90)
>>   0 inform,   0 warnings,   1 severes, 0 fatal for generate_erfc
>> PGF90-S-0038-Symbol, erfc, has not been explicitly declared
>> (generate_function.F90)
>>   0 inform,   0 warnings,   1 severes, 0 fatal for generate_graderfc
>> PGF90-S-0038-Symbol, erf, has not been explicitly declared
>> (generate_function.F90)
>>   0 inform,   0 warnings,   1 severes, 0 fatal for erfcvolume
>>
>>
>> To made it compile that part, I had to made the following small bugfix to
>> the
>> Module/generate_function.f90
>>
>> 550a551
>> >       REAL( DP)                 :: erfc
>> 666a668
>> >       REAL( DP )                :: erfc
>> 916a919
>> >     REAL( DP )  :: erf
>>
>> You might be interested to add this small bug fix to the trunk.
>>
>> Best,
>>
>> Samuel Ponce
>>
>>
>>
>> _______________________________________________
>> Q-e-developers mailing list
>> Q-e-developers at qe-forge.org
>> http://qe-forge.org/mailman/listinfo/q-e-developers
>>
>>
>
>
> --
> Paolo Giannozzi, Dept. Chemistry&Physics&Environment,
> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> Phone +39-0432-558216, fax +39-0432-558222
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20150815/906b1747/attachment.html>