[CP2K:8565] error in qs_ks_methods.F depending on number of cores (intel compiler)

Iain Bethune i.be... at epcc.ed.ac.uk
Fri Jan 27 14:37:06 UTC 2017


Further to what I reported this morning, I found that the MKL release 2017.1.132 is buggy.  I was able to get a successful build using the compiler from the 2017.1.132 release and MKL from the previous (2017.0.098) release.

- Iain

--

Iain Bethune
Project Manager, EPCC

Email: i.be... at epcc.ed.ac.uk
Twitter: @IainBethune @PrimeGrid @CP2Kproject
Web: http://www2.epcc.ed.ac.uk/~ibethune
Tel/Fax: +44 (0)131 651 7183/6555
Mob: +44 (0)7598317015
Addr: 2404 JCMB, The King's Buildings, Peter Guthrie Tait Road, Edinburgh, EH9 3FD

> On 27 Jan 2017, at 11:04, Iain Bethune <i.be... at epcc.ed.ac.uk> wrote:
> 
> Hi Laurent,
> 
> I got hold of the 2017.1.132 release and ran a regtest of the latest trunk code.  Barring one regtest with numerical problems, it ran fine, so I don’t think there is something fundamentally broken with that version.  Could you send your input file, and try running with the GLOBAL%PRINT_LEVEL HIGH so some more information is printed out before the error occurs?  Maybe someone can spot the problem.  Did you have any success running this input with previous builds of CP2K?
> 
> - Iain
> --
> 
> Iain Bethune
> Project Manager, EPCC
> 
> Email: i.be... at epcc.ed.ac.uk
> Twitter: @IainBethune @PrimeGrid @CP2Kproject
> Web: http://www2.epcc.ed.ac.uk/~ibethune
> Tel/Fax: +44 (0)131 651 7183/6555
> Mob: +44 (0)7598317015
> Addr: 2404 JCMB, The King's Buildings, Peter Guthrie Tait Road, Edinburgh, EH9 3FD
> 
>> On 19 Jan 2017, at 17:56, Laurent Joly <ljoly... at gmail.com> wrote:
>> 
>> Hi Iain,
>> 
>> Thank you very much for your fast answer.
>> 
>> I tried a simplified version of the arch file you sent (enclosed
>> raptor2.psmp), but I keep getting the same behavior. I also tried to
>> compile all the files with -O1 (enclosed raptor1.psmp and
>> Linux-x86-64-intel-mic.psmp), or even -O0 (enclosed raptor3.psmp), but
>> here again it makes no difference.
>> 
>> When I try to run the regtests, the system hangs after printing
>> "Copying tests into working directory ... done!"... But possibly this
>> is another problem...
>> 
>> Best regards,
>> Laurent
>> 
>> 
>> 2017-01-19 14:57 GMT+00:00 Iain Bethune <i.be... at epcc.ed.ac.uk>:
>>> Hi Laurent,
>>> 
>>> We don’t have experience yet with exactly that version (ifort 17.0.1), but with the 17.0.0 compiler we find that a couple of files need to be built at reduced optimisation level - see https://dashboard.cp2k.org/archive/epcc-phi-psmp/rev_17693.txt for an example arch file which is working.
>>> 
>>> Were you able to run the CP2K regression tests with success?
>>> 
>>> Cheers
>>> 
>>> - Iain
>>> 
>>> --
>>> 
>>> Iain Bethune
>>> Project Manager, EPCC
>>> 
>>> Email: i.be... at epcc.ed.ac.uk
>>> Twitter: @IainBethune @PrimeGrid @CP2Kproject
>>> Web: http://www2.epcc.ed.ac.uk/~ibethune
>>> Tel/Fax: +44 (0)131 651 7183/6555
>>> Mob: +44 (0)7598317015
>>> Addr: 2404 JCMB, The King's Buildings, Peter Guthrie Tait Road, Edinburgh, EH9 3FD
>>> 
>>>> On 19 Jan 2017, at 14:51, Laurent Joly <ljoly... at gmail.com> wrote:
>>>> 
>>>> Dear all,
>>>> 
>>>> I've been trying to compile CP2K 4.1 (I also tried with the 3.0 version) using the intel compiler version 2017.1.132.
>>>> 
>>>> I used the Linux-x86-64-intel-host.popt and Linux-x86-64-intel-host.psmp arch files, and I had to modify the Linux-x86-64-intel-mic.psmp file following the instructions of the intel mkl link line advisor:  https://software.intel.com/en-us/articles/intel-mkl-link-line-advisor to make it work:
>>>> 
>>>> $ diff Linux-x86-64-intel-mic.psmp.backup Linux-x86-64-intel-mic.psmp
>>>> 231d230
>>>> <             $(MKLROOT)/lib/intel64/libmkl_core.a \
>>>> 233,234c232,234
>>>> <           -Wl,--end-group \
>>>> <           $(MKLROOT)/lib/intel64/libmkl_blacs_intelmpi_lp64.a
>>>> ---
>>>>>           $(MKLROOT)/lib/intel64/libmkl_core.a \
>>>>>           $(MKLROOT)/lib/intel64/libmkl_blacs_intelmpi_lp64.a \
>>>>>         -Wl,--end-group
>>>> 
>>>> Then I had to run: ulimit -s unlimited in order to get rid of segmentation faults, and finally it worked... on 16 cores, but as soon as I tried to increase the number of cores I got this error message:
>>>> 
>>>> KS energy is an abnormal value (NaN/Inf).
>>>> qs_ks_methods.F:785
>>>> 
>>>> I enclose the complete output. Do you have some ideas on what could be the problem?
>>>> 
>>>> Best regards,
>>>> Laurent
>>>> 
>>>> 
>>>> --
>>>> You received this message because you are subscribed to the Google Groups "cp2k" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com.
>>>> To post to this group, send email to cp... at googlegroups.com.
>>>> Visit this group at https://groups.google.com/group/cp2k.
>>>> For more options, visit https://groups.google.com/d/optout.
>>>> <out>
>>> 
>>> 
>>> --
>>> The University of Edinburgh is a charitable body, registered in
>>> Scotland, with registration number SC005336.
>>> 
>>> --
>>> You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
>>> To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/C2FO7VAUZEg/unsubscribe.
>>> To unsubscribe from this group and all its topics, send an email to cp2k+uns... at googlegroups.com.
>>> To post to this group, send email to cp... at googlegroups.com.
>>> Visit this group at https://groups.google.com/group/cp2k.
>>> For more options, visit https://groups.google.com/d/optout.
>> 
>> 
>> 
>> -- 
>> Laurent JOLY
>> 
>> Institut Lumière Matière - Université Lyon 1
>> Campus de la Doua, bâtiment Brillouin
>> 43 bd du 11 novembre 1918
>> 69622 Villeurbanne Cedex, France
>> 
>> Tél: 04 724 326 11 - Fax: 04 724 326 48
>> Web: http://ilm-perso.univ-lyon1.fr/~ljoly/
>> 
>> -- 
>> You received this message because you are subscribed to the Google Groups "cp2k" group.
>> To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com.
>> To post to this group, send email to cp... at googlegroups.com.
>> Visit this group at https://groups.google.com/group/cp2k.
>> For more options, visit https://groups.google.com/d/optout.
>> <Linux-x86-64-intel-mic.psmp><raptor1.psmp><raptor2.psmp><raptor3.psmp>
> 


-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.




More information about the CP2K-user mailing list