[CP2K:10047] Re: libint-2.2.2 error when building CP2K 5.1 with Intel Composer 2017 update5

Alfio Lazzaro alfio.... at gmail.com
Thu Mar 1 13:12:40 UTC 2018


OK, looking at -O0, you can ignore the TMC/regtest_ana_post_proc (they are 
unstable with the number of ranks). Same for 
the QS/regtest-ot-1-vib and regtest-optbas/opt-1.inp.
You can skip these tests by adding:

-skipdirs TMC/regtest_ana_post_proc -skipdirs QS/regtest-ot-1-vib

to your do_regtest command. 

For the remaining test, somehow you are using an old reference. For the 
CP2K dashboard (http://cp2k-www.epcc.ed.ac.uk/phi/popt/regtest-0 ) I see 
the following values:

regtest-3-4/Al2003.inp ==> -202.455703790495818

which very close to your value. Therefore I would conclude that -O0 build 
passes all tests.
For the -O2, I see other 4 tests failing:

Fist/regtest-7-2/UO2-4x4x4-cs-fixd-npt.inp
Fist/regtest-1-4/water_atprop_ewald.inp
Fist/regtest-4/ethene-no-restraint.inp
Fist/regtest-4/ethene-no-restraint.inp

The last 3 are off by small differences. The first one is not clear... I 
don't know if it is worth to investigate more... I would leave as it is...

Alfio


Il giorno giovedì 1 marzo 2018 13:45:54 UTC+1, Jerry Tanoury ha scritto:
>
> Here is how I run my regtests
>
> tools/regtesting/do_regtest -arch CP2K_allintel -version popt -nosvn 
> -nobuild -maxtasks 4.  Attached are the error files for the -O2 
> (error_summary-O2) and -O0 (error_summary-O0) optimizations, fyi.  And to 
> be explicitly clear, my FCFLAGS (for the -O2 build) were as follows:
>
> FCFLAGS  = $(DFLAGS) $(INCS) -O2 -fp-model source -fp-model precise -g 
> -traceback -fpp -free -xHost \
>            -diag-disable $(DIAG_DISABLE)
>
>
>
>
>
> On Thu, Mar 1, 2018 at 7:10 AM, Alfio Lazzaro <alfi... at gmail.com 
> <javascript:>> wrote:
>
>> Hi Jerry,
>> That's a good news. BTW, how many ranks and threads are you using? The 
>> regtests references are for 2 ranks + 2 threads, therefore you can expect 
>> some wrong values (with somehow reasonable differences) if you use a 
>> different number of ranks/threads.
>>
>> Alfio
>>
>> Il giorno giovedì 1 marzo 2018 13:05:53 UTC+1, Jerry Tanoury ha scritto:
>>>
>>> Hi Alfio,
>>> I figured it out.  During the LIBINT build, configure was not reading my 
>>> values for libint-max-am and libderiv-max-am1.  When I removed all compiler 
>>> flags and configured with:
>>> --prefix=/cluster/home/tanoury/CP2K/intelbuilt_packages/2017u5/libint-1.1.4 
>>> CC=icc CXX=icpc --with-libint-max-am=6 --with-libderiv-max-am1=5 everything 
>>> worked as expected, and the CP2K build completed without warnings.  The 
>>> regtest gave 364 WRONG, but when I added "-fp-model source -fp-model 
>>> precise", and kept -O2 optimization level,  I got 12 WRONG 0 FAILED.  
>>> Replacing -O2 with -O0 (and keeping the -fp-model flags) gave 8 WRONG 0 
>>> FAILED.
>>>
>>> Best regards,
>>> Jerry 
>>>
>>> On Wed, Feb 28, 2018 at 10:42 AM, Jerry Tanoury <jerr... at gmail.com> 
>>> wrote:
>>>
>>>> Hi Alfio,
>>>> Thank you for the reply.  I tried your suggestion, but the warnings 
>>>> were still there.
>>>>
>>>> On further investigation, the problem seems to be from building 
>>>> libint-1.1.4 with Intel compilers.  When I use a GNU-built version of 
>>>> libint (GCC 4.8.5), I do not get the error.  I am running regtest now 
>>>> (getting several wrong results, but no runtime failures).  
>>>>
>>>> I am curious, how were the dependencies built for  
>>>> Linux-x86-64-intel.popt 
>>>> <https://dashboard.cp2k.org/archive/epcc-phi-popt/index.html> in the 
>>>> Dashboard.  Intel or GNU compilers?  I am using that arch file as my 
>>>> starting point.
>>>>
>>>> Best regards,
>>>> Jerry
>>>>
>>>> On Wed, Feb 28, 2018 at 2:55 AM, Alfio Lazzaro <alfi... at gmail.com> 
>>>> wrote:
>>>>
>>>>> Hi Jerry,
>>>>> OK, thanks for your clarification. Then I think the problem is that 
>>>>> LIBINT uses values starting at 0, while CP2K starts at 1.
>>>>> I mean, you have to use the following values to make things consistent:
>>>>>
>>>>> LIBINT: --with-libint-max-am=5 --with-libderiv-max-am1=4
>>>>> CP2K: -D__LIBINT_MAX_AM=6 -D__LIBDERIV_MAX_AM1=5
>>>>>
>>>>> Alfio
>>>>>
>>>>>
>>>>> Il giorno mercoledì 28 febbraio 2018 01:21:34 UTC+1, Jerry Tanoury ha 
>>>>> scritto:
>>>>>>
>>>>>> Hi Alfio,
>>>>>> My mistake, I am using libint-1.1.4 (I accidentally wrote the version 
>>>>>> for libxc).
>>>>>> .
>>>>>> On Tuesday, February 27, 2018 at 5:21:41 PM UTC-5, Alfio Lazzaro 
>>>>>> wrote:
>>>>>>>
>>>>>>> Dear Jerry,
>>>>>>> As far as I know CP2K doesn't support libint-2...
>>>>>>>
>>>>>>> Alfio
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Il giorno martedì 27 febbraio 2018 19:46:25 UTC+1, Jerry Tanoury ha 
>>>>>>> scritto:
>>>>>>>>
>>>>>>>> Hi,
>>>>>>>> I am trying to build CP2K 5.1 with Intel compilers and MKL libs.  I 
>>>>>>>> am using ComposerXE 2017 update 5 (2017.5.239) with associated MKL libs, 
>>>>>>>> and MPI library parallel studio 2017.5.061.  My arch file is attached.  At 
>>>>>>>> the end of the build I get the following error:
>>>>>>>>
>>>>>>>>
>>>>>>>> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>>>>>>>> ld: ld: Warning: size of symbol `build_deriv1_eri' changed from 
>>>>>>>> 5000 in 
>>>>>>>> /cluster/home/tanoury/CP2K/cp2k-5.1/lib/CP2K_allintel/popt/libcp2kmain.a(hfx_libint_wrapper.o) 
>>>>>>>> to 10368 in 
>>>>>>>> /cluster/home/tanoury/CP2K/intelbuilt_packages/2017u5/libint-1.1.4/lib/libderiv.a(init_libderiv.o)Warning: 
>>>>>>>> size of symbol `build_deriv1_eri' changed from 5000 in 
>>>>>>>> /cluster/home/tanoury/CP2K/cp2k-5.1/lib/CP2K_allintel/popt/libcp2kmain.a(hfx_libint_wrapper.o) 
>>>>>>>> to 10368 in 
>>>>>>>> /cluster/home/tanoury/CP2K/intelbuilt_packages/2017u5/libint-1.1.4/lib/libderiv.a(init_libderiv.o)
>>>>>>>>
>>>>>>>> ld: Warning: size of symbol `build_deriv1_eri' changed from 5000 in 
>>>>>>>> /cluster/home/tanoury/CP2K/cp2k-5.1/lib/CP2K_allintel/popt/libcp2kmain.a(hfx_libint_wrapper.o) 
>>>>>>>> to 10368 in 
>>>>>>>> /cluster/home/tanoury/CP2K/intelbuilt_packages/2017u5/libint-1.1.4/lib/libderiv.a(init_libderiv.o)
>>>>>>>> ld: ld: Warning: size of symbol `build_eri' changed from 10368 in 
>>>>>>>> /cluster/home/tanoury/CP2K/cp2k-5.1/lib/CP2K_allintel/popt/libcp2kmain.a(hfx_libint_wrapper.o) 
>>>>>>>> to 19208 in 
>>>>>>>> /cluster/home/tanoury/CP2K/intelbuilt_packages/2017u5/libint-1.1.4/lib/libint.a(init_libint.o)Warning: 
>>>>>>>> size of symbol `build_eri' changed from 10368 in 
>>>>>>>> /cluster/home/tanoury/CP2K/cp2k-5.1/lib/CP2K_allintel/popt/libcp2kmain.a(hfx_libint_wrapper.o) 
>>>>>>>> to 19208 in 
>>>>>>>> /cluster/home/tanoury/CP2K/intelbuilt_packages/2017u5/libint-1.1.4/lib/libint.a(init_libint.o)
>>>>>>>>
>>>>>>>> ld: Warning: size of symbol `build_eri' changed from 10368 in 
>>>>>>>> /cluster/home/tanoury/CP2K/cp2k-5.1/lib/CP2K_allintel/popt/libcp2kmain.a(hfx_libint_wrapper.o) 
>>>>>>>> to 19208 in 
>>>>>>>> /cluster/home/tanoury/CP2K/intelbuilt_packages/2017u5/libint-1.1.4/lib/libint.a(init_libint.o)
>>>>>>>>
>>>>>>>> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>>>>>>>>
>>>>>>>> My build of libint-2.2.2 is: ../configure 
>>>>>>>> --prefix=/cluster/home/tanoury/CP2K/intelbuilt_packages/2017u5/libint-1.1.4 
>>>>>>>> --with-libint-max-am=6 --with-libderiv-max-am1=5 --with-cc="icc -O2 -xHost" 
>>>>>>>> --with-cc-optflags="-O2 -xHost" --with-cxx="icpc -O2 -xHost" 
>>>>>>>> --with-cxx-optflags="-O2 -xHost"
>>>>>>>>
>>>>>>>> I've read previous threads about matching the am and am1 values 
>>>>>>>> between the CP2K build and the libint build, which I have done.  I'm not 
>>>>>>>> sure what's going on here.  
>>>>>>>>
>>>>>>>> Thank you for the help,
>>>>>>>> Jerry
>>>>>>>>
>>>>>>> -- 
>>>>> 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+... 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.
>>>>>
>>>>
>>>>
>>> -- 
>> 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+... at googlegroups.com <javascript:>.
>> To post to this group, send email to cp... at googlegroups.com <javascript:>
>> .
>> Visit this group at https://groups.google.com/group/cp2k.
>> For more options, visit https://groups.google.com/d/optout.
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180301/158ad0af/attachment.htm>


More information about the CP2K-user mailing list