[Wien] Multicore on i7 iMac
Peter Blaha
pblaha at theochem.tuwien.ac.at
Fri Apr 19 08:32:42 CEST 2013
From this message alone, we one cannot say anything, maybe except that
the calculations seem to have diverged (large CTEST).
But when starting from a converged calculation, this is very unusual.
PS: In a "slab" calculation it is almost always possible to setup the
slab such that it has inversion symmetry and thus you would have
case.in1 instead of case.in1c.
Inversion for slabs is a) much faster, b) avoids spurious dipoles from
two different surfaces.
On 04/17/2013 11:03 PM, pluto at physics.ucdavis.edu wrote:
> Dear Prof. Blaha and WIEN2k experts,
>
> I have 4 physical cores (Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz). It
> seems that on my compilation using HT and filling up all 8 threads makes
> some particular calculation just a bit faster compared to the settings you
> have suggested, but with HT CPU gets more hot (fan is on more often), so
> it makes no sense. I will use the settings you have recommended.
>
> I have now an error for the slab with spin-polarized but without spin
> orbit (see below part of STDOUT file). I tried to look at old emails from
> this group, but could not quickly find a solution. Same slab has converged
> before in a non-parallel mode with spin-polarized and with spin-orbit. I
> use cutoff 8 Ry and:
>
> K-VECTORS FROM UNIT:4 -11.0 5.5 933 emin/emax/nband
>
> in case.in1c.
>
> Regards,
> Lukasz
>
>
> ...
> CORE END
> CORE END
> MIXER END
> in cycle 22 ETEST: .3816818050000000 CTEST: 1.9705727
> LAPW0 END
> LAPW1 END
> LAPW1 END
> LAPW1 END
> LAPW1 END
> LAPW2 - FERMI; weighs written
> LAPW2 END
> LAPW2 END
> SUMPARA END
> LAPW2 - FERMI; weighs written
> LAPW2 END
> LAPW2 END
> SUMPARA END
> CORE END
> CORE END
> MIXER END
> in cycle 23 ETEST: .2942263650000000 CTEST: 2.3441252
> LAPW0 END
> LAPW1 END
> LAPW1 END
> LAPW1 END
> LAPW1 END
> FERMI - Error
> cp: .in.tmp: No such file or directory
>
>> stop error
>
>
>
>
> How many "real" cores do you have ? Most likely only 4 (the 8 comes from
> hyperthreading, but for numerical intensive application one should
> probably not use hyperthreading).
>
> So the "best" performance can probably be reached either by:
>
> OMP_NUM_THREADS=2 and 2 lines in .machines or
> OMP_NUM_THREADS=1 and 4 lines in .machines
>
> (it may even depend on the number of k-points in the specific case ..)
>
>
>
> On 04/16/2013 02:49 PM, pluto at physics.ucdavis.edu wrote:
>> Dear Prof. Blaha,
>>
>> Thank you for the answer. In the meantime I have realized this mistake.
>>
>> I have now all 8 threads practically fully utilized (HT Intel i7 in iMac)
>> for lapw1 and lapw2. It reduced the iteration from approx. 7.2min to
>> 5.5min (compared to utilizing 4 threads only with OMP_NUM_THREADS).
>>
>> I think it solves my problems for now. Again thank you for your support
>> and rapid answers.
>>
>> Regards,
>> Lukasz
>>
>>
>>
>>> Your .machines file is wrong. It contains more than one hostname per
>>> line (or has a localhost:2)
>>>
>>> With the proper .machines file, mpirun is not needed:
>>> > bash: mpirun: command not found
>>>
>>> -------------------- .machines --------------
>>> 1:localhost
>>> 1:localhost
>>>
>>> This file with split the klist into two parts and run two lapw1-jobs
>>> simultaneously.
>>>
>>> On 04/16/2013 11:48 AM, pluto at physics.ucdavis.edu wrote:
>>>> Hello Prof. Blaha, Prof. Marks,
>>>>
>>>> ssh localhost works now without login!!
>>>>
>>>> I have more errors now when trying to run parallel mode, see below.
>>>>
>>>> In UG there are sections 5.5.1 (k-point parallelization) and 5.5.2 (MPI
>>>> parallelization). I understand these two modes are separate, and I would
>>>> like to focus on k-point parallelization for now. I am not sure why
>>>> there
>>>> is an error regarding the mpirun. My parallel_options file is now:
>>>>
>>>> setenv USE_REMOTE 0
>>>> setenv MPI_REMOTE "1"
>>>> setenv WIEN_GRANULARITY 1
>>>>
>>>> But with other options I have the same error.
>>>>
>>>> I would appreciate if there is something obvious which I do wrong. In
>>>> any
>>>> case I will continue to work on the issue with my IT department here.
>>>>
>>>> Regards,
>>>> Lukasz
>>>>
>>>>
>>>>
>>>> LAPW0 END
>>>> bash: mpirun: command not found
>>>>
>>>> real 0m0.001s
>>>> user 0m0.000s
>>>> sys 0m0.000s
>>>> Mo-bulk-so.scf1_1: No such file or directory.
>>>> ERROR IN OPENING UNIT: 9
>>>> FILENAME:
>>>> ./Mo-bulk-so.vector_1
>>>> STATUS: old FORM:unformatted
>>>> OPEN FAILED
>>>>
>>>>
>>>>
>>>>
>>>> -------- Original Message --------
>>>> Subject: Re: [Wien] Multicore on i7 iMac
>>>> Date: Mon, 15 Apr 2013 08:49:39 -0500
>>>> From: Laurence Marks <L-marks at northwestern.edu>
>>>> Reply-To: A Mailing list for WIEN2k users
>>>> <wien at zeus.theochem.tuwien.ac.at>
>>>> To: A Mailing list for WIEN2k users <wien at zeus.theochem.tuwien.ac.at>
>>>>
>>>> You may also be able to turn off USE_REMOTE and MPI_REMOTE (set both
>>>> to 0) and/or use something other than ssh to launch processes.
>>>>
>>>> On Mon, Apr 15, 2013 at 8:33 AM, Peter Blaha
>>>> <pblaha at theochem.tuwien.ac.at> wrote:
>>>>> Try it again. I think it ask this disturbing question only once !
>>>>>
>>>>> otherwise: you must be able to do:
>>>>>
>>>>> ssh localhost
>>>>>
>>>>> and login without any other response.
>>>>>
>>>>>> The authenticity of host 'localhost (::1)' can't be established.
>>>>>> RSA key fingerprint is
>>>>>> 50:c3:da:fa:0c:35:c5:aa:d1:b4:c1:52:a1:18:08:c2.
>>>>>> Are you sure you want to continue connecting (yes/no)? yes
>>>>>>
>>>>>> ^C
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> -------- Original Message --------
>>>>>> Subject: Re: [Wien] Multicore on i7 iMac
>>>>>> Date: Mon, 15 Apr 2013 08:16:11 +0200
>>>>>> From: Peter Blaha <pblaha at theochem.tuwien.ac.at>
>>>>>> Reply-To: A Mailing list for WIEN2k users
>>>> <wien at zeus.theochem.tuwien.ac.at>
>>>>>> To: A Mailing list for WIEN2k users
>>>>>> <wien at zeus.theochem.tuwien.ac.at>
>>>>>>
>>>>>> As you could see from your "top" command, only 1 core is used.
>>>>>>
>>>>>> The "simplest" thing is to set:
>>>>>>
>>>>>> export OMP_NUM_THREADS=2 (or 4) (a commented line is already in
>>>>>> your
>>>>>> .bashrc after "userconfig_lapw")
>>>>>>
>>>>>> This will use 2 (4) cores for parts of WIEN2k which uses the
>>>>>> mkl-library.
>>>>>> ---------------------------
>>>>>> Next is k-parallel mode (see UG for description), where you can use
>>>>>> all
>>>>>> your cores.
>>>>>>
>>>>>> We also have mpi-parallel, but I would not recommend it for a single
>>>>>> mac,
>>>>>> unless you have a problem with just one k-point.
>>>>>>
>>>>>> Please also notice the recent posting on the mailing-list about a
>>>>>> recommended compiler option for a Mac (-heap-arrays), otherwise you
>>>>>> cannot
>>>>>> run wien2k on larger systems
>>>>>>
>>>>>> Peter Blaha
>>>>>>
>>>>>>
>>>>>>
>>>>>> From:pluto at physics.ucdavis.edu
>>>>>> Date:04/14/2013 07:48 PM
>>>>>> To:"A Mailing list for WIEN2k users" <wien at zeus.theochem.tuwien.ac.at>
>>>>>>
>>>>>> Hello WIEN2k experts,
>>>>>>
>>>>>> I have a very simple question.
>>>>>>
>>>>>> Do I need to edit the .machines file for the multicore operation of
>>>>>> the
>>>>>> Intel i7 Quad Core CPU?
>>>>>>
>>>>>> My IT department (in FZ Juelich, Germany) has helped to compile Wien2k
>>>>>> on
>>>>>> an iMac with i7 CPU. It works very nice, no problem to calculate 15
>>>>>> layer
>>>>>> slab. However, I have a feeling, that all is done on a single core,
>>>>>> and
>>>>>> this is a real waste to time. I attach the screenshot of "top" program
>>>>>> under the terminal, with lapw1c doing 100 k-points for band structure.
>>>>>>
>>>>>> Regards,
>>>>>> Lukasz
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> Wien mailing list
>>>>>> Wien at zeus.theochem.tuwien.ac.at
>>>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>>> SEARCH the MAILING-LIST at:
>>>>>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> Wien mailing list
>>>>>> Wien at zeus.theochem.tuwien.ac.at
>>>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>>> SEARCH the MAILING-LIST at:
>>>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>>>>>
>>>>>
>>>>> --
>>>>>
>>>>> P.Blaha
>>>>> --------------------------------------------------------------------------
>>>>> Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
>>>>> Phone: +43-1-58801-165300 FAX: +43-1-58801-165982
>>>>> Email: blaha at theochem.tuwien.ac.at WWW:
>>>>> http://info.tuwien.ac.at/theochem/
>>>>> --------------------------------------------------------------------------
>>>>> _______________________________________________
>>>>> Wien mailing list
>>>>> Wien at zeus.theochem.tuwien.ac.at
>>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>> SEARCH the MAILING-LIST at:
>>>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>>>
>>>>
>>>>
>>>> -- Professor Laurence Marks Department of Materials Science and
>>>> Engineering Northwestern University www.numis.northwestern.edu
>>>> 1-847-491-3996 "Research is to see what everybody else has seen, and to
>>>> think what nobody else has thought" Albert Szent-Gyorgi
>>>> _______________________________________________ Wien mailing list
>>>> Wien at zeus.theochem.tuwien.ac.at
>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the
>>>> MAILING-LIST at:
>>>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>>>
>>>> _______________________________________________
>>>> Wien mailing list
>>>> Wien at zeus.theochem.tuwien.ac.at
>>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>> SEARCH the MAILING-LIST at:
>>>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>>>
>>>
>>> --
>>>
>>> P.Blaha
>>> --------------------------------------------------------------------------
>>> Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
>>> Phone: +43-1-58801-165300 FAX: +43-1-58801-165982
>>> Email: blaha at theochem.tuwien.ac.at WWW:
>>> http://info.tuwien.ac.at/theochem/
>>> --------------------------------------------------------------------------
>>> _______________________________________________
>>> Wien mailing list
>>> Wien at zeus.theochem.tuwien.ac.at
>>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>> SEARCH the MAILING-LIST at:
>>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>>
>>
>>
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> SEARCH the MAILING-LIST at:
> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>
>
--
P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300 FAX: +43-1-58801-165982
Email: blaha at theochem.tuwien.ac.at WWW:
http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------
More information about the Wien
mailing list