[Wien] LAPW2 looping indefinitely

Roberto Iglesias roberto.iglesias at psi.ch
Mon Jun 19 17:08:52 CEST 2006


Problem solved!

I did finally as Torsten Andersen suggested, copied the binaries from 
the server to my PC and now for large k-meshes LAPW2 does not get stuck 
and the calculation reaches convergence. This is just a workaround, 
however, and I still don't get why it did not work with my previous 
installation. But anyway, thanks a lot for your help
Roberto

Roberto Iglesias wrote:
> Dear all,
> 
> sorry for being back again with this problem. I'm pretty sure it is not 
> a compiler problem, but it may be a library one. How can I recompile 
> lapack_lapw with a lower optimization, as Torsten suggested in his 
> previous answer?
> And also: instead of using the binaries from the installation that 
> works, which is not on a PC, but on a central server, wouldn't it be 
> better to use the pre-compiled WIEN2k executables?
> And finally: I cannot use other compiler different from pgf, because it 
> is the one supported at our institute and I don't even have the proper 
> mkl libraries installed on my PC. But in the server I mentioned the 
> compiler and library settings are exactly the same and the system 
> terminates the SCF cycle without indefinitely looping in lapw2.
> 
> Thanks for your help!
> 
> Roberto
> 
> Roberto Iglesias wrote:
> 
>>Just the recommended options for linuxpgi:
>>
>>Recommended options for system linuxpgi are:
>>         Compiler options:        -Mfreeform -fast -Kieee
>>         Linker Flags:            -L../SRC_lib
>>         Preprocessor flags:      '-DParallel'
>>         R_LIB (LAPACK+BLAS):     -llapack_lapw -lblas
>>
>>   Current settings:
>>     O   Compiler options:        -Mfreeform -fast -Kieee
>>     L   Linker Flags:            -L../SRC_lib
>>     P   Preprocessor flags       '-DParallel'
>>     R   R_LIB (LAPACK+BLAS):     -llapack_lapw -lblas
>>
>>In the second system I've just mentioned these were exactly the same...
>>
>>Thanks for your answers!
>>
>>Roberto
>>
>>
>>
>>
>>C J Kenneth Tan -- OptimaNumerics wrote:
>>
>>
>>>Roberto,
>>>
>>>Which implementation of LAPACK are you using? 
>>>
>>>Kenneth Tan
>>>-----------------------------------------------------------------------
>>>C J Kenneth Tan, PhD
>>>OptimaNumerics Ltd                    Telephone: +44 798 941 7838
>>>E-mail: cjtan at OptimaNumerics.com      Telephone: +44 207 099 4428
>>>Web: http://www.OptimaNumerics.com    Facsimile: +44 207 100 4572
>>>-----------------------------------------------------------------------
>>>
>>>On 2006-06-02 11:04 +0200 Torsten Andersen (thor at physik.uni-kl.de) wrote:
>>>
>>>
>>>
>>>>Date: Fri, 02 Jun 2006 11:04:30 +0200
>>>>From: Torsten Andersen <thor at physik.uni-kl.de>
>>>>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>
>>>>Subject: Re: [Wien] LAPW2 looping indefinitely
>>>>
>>>>Dear Roberto,
>>>>
>>>>I have seen this problem with an IBM compiler (on IBM p630 and p690). 
>>>>It is
>>>>most likely a compiler problem or a library problem related to one of 
>>>>the
>>>>LAPACK routines. If you use lapack_lapw, try to compile it with lower
>>>>optimization, and if this is not enough, do the same with lapw2.
>>>>
>>>>If you suspect it is not a compiler problem, try to use the binaries 
>>>>from the
>>>>PC that works on the one that does not.
>>>>
>>>>The pgf compiler is btw not the most bugfree compiler around... try 
>>>>ifc (also
>>>>on AMD).
>>>>
>>>>Best regards,
>>>>Torsten Andersen.
>>>>
>>>>Roberto Iglesias wrote:
>>>>
>>>>
>>>>>Dear all
>>>>>
>>>>>When running some tests in my own Desktop PC I find the following 
>>>>>problem.
>>>>>Whenever I use a number of k points larger than about 2000, 
>>>>>regardless of
>>>>>the system (pure element, 2 or 4 atom unit cell, 2x2x2 supercell or 
>>>>>3x3x3
>>>>>supercell), the calculation gets stuck min LAPW2 in the first scf 
>>>>>cycle,
>>>>>with no error messages.
>>>>>STDOUT looks like
>>>>>
>>>>>LAPW0 END
>>>>>LAPW1 END
>>>>>LAPW1 END
>>>>>
>>>>>and case.dayfile:
>>>>>
>>>>>   start       (Thu Jun  1 10:22:26 CEST 2006) with lapw0 (200/40 to 
>>>>>go)
>>>>>
>>>>>   cycle 1     (Thu Jun  1 10:22:26 CEST 2006)         (200/40 to go)
>>>>>
>>>>>
>>>>>>  lapw0       (10:22:26) 3.114u 0.058s 0:04.33 72.9%  0+0k 0+0io 
>>>>>
>>>>>0pf+0w
>>>>>
>>>>>>  lapw1  -up          (10:22:30) 44.132u 1.877s 0:50.90 90.3% 0+0k 
>>>>>
>>>>>0+0io
>>>>>0pf+0w
>>>>>
>>>>>>  lapw1  -dn          (10:23:21) 44.175u 1.753s 0:49.45 92.8% 0+0k 
>>>>>
>>>>>0+0io
>>>>>0pf+0w
>>>>>
>>>>>>  lapw2 -up   (10:24:11)
>>>>>
>>>>>
>>>>>The system has 2 GB RAM, a P4 with 3.2 GHz processor, a 200 GB hard 
>>>>>disk,
>>>>>the WIEN2k version is 06.2, the operating system is a derivation of
>>>>>Scientific Linux 4 (SL42 version 0.30) and the fortran 90 compiler 
>>>>>is pgf90.
>>>>>The same calculation in an older PC runs smoothly. I could not find 
>>>>>anything
>>>>>similar to this in the FAQ or the mailing list.
>>>>>Any suggestions?
>>>>>Thanks in advance
>>>>>Roberto Iglesias
>>>>>_______________________________________________
>>>>>Wien mailing list
>>>>>Wien at zeus.theochem.tuwien.ac.at
>>>>>http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>>
>>>>
>>>>-- 
>>>>Dr. Torsten Andersen        TA-web: http://deep.at/myspace/
>>>>AG H�bner, Department of Physics, Kaiserslautern University
>>>>http://cmt.physik.uni-kl.de    http://www.physik.uni-kl.de/
>>>>_______________________________________________
>>>>Wien mailing list
>>>>Wien at zeus.theochem.tuwien.ac.at
>>>>http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>>>
>>>>
>>>>
>>>>------------------------------------------------------------------------
>>>>
>>>>_______________________________________________
>>>>Wien mailing list
>>>>Wien at zeus.theochem.tuwien.ac.at
>>>>http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>
>>
>>_______________________________________________
>>Wien mailing list
>>Wien at zeus.theochem.tuwien.ac.at
>>http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien


More information about the Wien mailing list