[Wien] LAPW2 crashed when running in parallel

Maxim Rakitin rms85 at physics.susu.ac.ru
Mon Nov 1 18:15:30 CET 2010


Hi Wei,

The parallel_options file manages how parallel programs run, so change 
the following line in it:

    setenv WIEN_MPIRUN "mpirun -np _NP_ -hostfile _HOSTS_ _EXEC_"

to

    setenv WIEN_MPIRUN "mpirun -np _NP_ -machinefile _HOSTS_ _EXEC_"

Your .machine0/1/2 files are correct,

Also I believe that 'USE_REMOTE' variable which is set to 1 makes 
parallel scripts (I mean lapw[012]para_lapw) to be launched using 
ssh/rsh. So switch it to '0'. I'm not sure about 'MPI_REMOTE' option, 
it's a new one. Try to set different values (0 or 1) for it.

Hope this will help.

Best regards,
    Maxim Rakitin
    email: rms85 at physics.susu.ac.ru
    web: http://www.susu.ac.ru


01.11.2010 21:35, Wei Xie пишет:
> Hi Maxim,
>
> Thanks for the follow-up!
>
> I think it should be -machinefile  that's appropriate. Here's the help:
> -machinefile                 # file mapping procs to machine
>
> No -hostfile option mentioned for my current version of MPI in the help.
>
> Yes, the machine0/1/2 files are exactly like what you described.
>
> The parallel_options is:
> setenv USE_REMOTE 1
> setenv MPI_REMOTE 1
> setenv WIEN_GRANULARITY 1
> setenv WIEN_MPIRUN "mpirun -np _NP_ -hostfile _HOSTS_ _EXEC_"
>
> I think the problem should be due to my MPI. However, even if disable 
> MPI parallelization, the problem still persists (no evident difference 
> in the output files, including case.dayfile, stdout and :log). Note we 
> can run with exactly the same set of input files in serial mode with 
> no problem.
>
> Again, thanks for your help!
>
> Cheers,
> Wei
>
>
> On Oct 31, 2010, at 11:27 PM, Maxim Rakitin wrote:
>
>> Dear Wei,
>>
>> Maybe -machinefile is ok for your mpirun. Which options are 
>> appropriate for it? What does help say?
>>
>> Try to restore your MPIRUN variable with -machinefile and rerun the 
>> calculation. Then see what is in .machine0/1/2 files and let us know. 
>> It should contain 8 lines of r1i0n0 node and 8 lines of r1i0n1 node.
>>
>> One more thing you should check is $WIENROOT/parallel_options file. 
>> What is its content?
>> Best regards,
>>     Maxim Rakitin
>>     email:rms85 at physics.susu.ac.ru
>>     web:http://www.susu.ac.ru
>>
>> 01.11.2010 9:06, Wei Xie пишет:
>>> Hi Maxim,
>>>
>>> Thanks for your reply!
>>> We tried MPIRUN=mpirun -np _NP_ -hostfile _HOSTS_ _EXEC_, but the 
>>> problem persists. The only difference is that stdout changes to 
>>> ''… MPI: invalid option -hostfile …''.
>>>
>>> Thanks,
>>> Wei
>>>
>>>
>>> On Oct 31, 2010, at 10:40 PM, Maxim Rakitin wrote:
>>>
>>>> Hi,
>>>>
>>>> It looks like Intel's mpirun doesn't have '-machinefile' option. 
>>>> Instead of this it has '-hostfile' option (form here: 
>>>> http://downloadmirror.intel.com/18462/eng/nes_release_notes.txt).
>>>>
>>>> Try 'mpirun -h' for information about options and apply appropriate.
>>>> Best regards,
>>>>     Maxim Rakitin
>>>>     email:rms85 at physics.susu.ac.ru
>>>>     web:http://www.susu.ac.ru
>>>>
>>>> 01.11.2010 4:56, Wei Xie пишет:
>>>>> Dear all WIEN2k community members:
>>>>>
>>>>> We encountered some problem when running in parallel 
>>>>> (K-point, MPI or both)--the calculations crashed at LAPW2. Note we 
>>>>> had no problem running it in serial. We have tried to diagnose the 
>>>>> problem, recompile the code with difference options and test with 
>>>>> difference cases and parameters based on similar problems reported 
>>>>> on the mail list, but the problem persists. So we write here 
>>>>> hoping someone can offer us some suggestion. We have attached 
>>>>> related files below for your reference. Your replies are 
>>>>> appreciated in advance!
>>>>>
>>>>> This is a TiC example running in both Kpoint and MPI parallel on 
>>>>> two nodes /r1i0n0/ and /r1i0n1/ (8cores/node):
>>>>>
>>>>> *1. **stdout **(abridged) *
>>>>> MPI: invalid option -machinefile
>>>>> real0m0.004s
>>>>> user0m0.000s
>>>>> sys0m0.000s
>>>>> ...
>>>>> MPI: invalid option -machinefile
>>>>> real0m0.003s
>>>>> user0m0.000s
>>>>> sys0m0.004s
>>>>> TiC.scf1up_1: No such file or directory.
>>>>>
>>>>> LAPW2 - Error. Check file lapw2.error
>>>>> cp: cannot stat `.in.tmp': No such file or directory
>>>>> rm: cannot remove `.in.tmp': No such file or directory
>>>>> *rm: cannot remove `.in.tmp1': No such file or directory*
>>>>> *
>>>>> *
>>>>> *2. TiC.dayfile (abridged) *
>>>>> ...
>>>>>     start (Sun Oct 31 16:25:06 MDT 2010) with lapw0 (40/99 to go)
>>>>>     cycle 1 (Sun Oct 31 16:25:06 MDT 2010) (40/99 to go)
>>>>>
>>>>> >   lapw0 -p(16:25:06) starting parallel lapw0 at Sun Oct 31 
>>>>> 16:25:07 MDT 2010
>>>>> -------- .machine0 : 16 processors
>>>>> invalid "local" arg: -machinefile
>>>>>
>>>>> 0.436u 0.412s 0:04.63 18.1%0+0k 2600+0io 1pf+0w
>>>>> >   lapw1  -up -p (16:25:12) starting parallel lapw1 at Sun Oct 31 
>>>>> 16:25:12 MDT 2010
>>>>> ->  starting parallel LAPW1 jobs at Sun Oct 31 16:25:12 MDT 2010
>>>>> running LAPW1 in parallel mode (using .machines)
>>>>> 2 number_of_parallel_jobs
>>>>>      r1i0n0 r1i0n0 r1i0n0 r1i0n0 r1i0n0 r1i0n0 r1i0n0 r1i0n0(1)   
>>>>>    r1i0n1 r1i0n1 r1i0n1 r1i0n1 r1i0n1 r1i0n1 r1i0n1 r1i0n1(1)     
>>>>>  r1i0n0 r1i0n0 r1i0n0 r1i0n0 r1i0n0 r1i0n0 r1i0n0 r1i0n0(1)   
>>>>>  Summary of lapw1para:
>>>>>    r1i0n0 k=0 user=0 wallclock=0
>>>>>    r1i0n1 k=0 user=0 wallclock=0
>>>>> ...
>>>>> 0.116u 0.316s 0:10.48 4.0%0+0k 0+0io 0pf+0w
>>>>> >   lapw2 -up -p (16:25:34) running LAPW2 in parallel mode
>>>>> **  LAPW2 crashed!
>>>>> 0.032u 0.104s 0:01.13 11.5%0+0k 82304+0io 8pf+0w
>>>>> error: command   /home/xiew/WIEN2k_10/lapw2para -up uplapw2.def   
>>>>> failed
>>>>>
>>>>> *3. uplapw2.error *
>>>>> Error in LAPW2
>>>>>  'LAPW2' - can't open unit: 18
>>>>>  'LAPW2' -        filename: TiC.vspup
>>>>>  'LAPW2' -          status: old          form: formatted
>>>>> **  testerror: Error in Parallel LAPW2
>>>>>
>>>>> *4. .machines*
>>>>> #
>>>>> 1:r1i0n0:8
>>>>> 1:r1i0n1:8
>>>>> lapw0:r1i0n0:8 r1i0n1:8
>>>>> granularity:1
>>>>> extrafine:1
>>>>>
>>>>> *5. compilers, MPI and options*
>>>>> Intel Compilers  and MKL 11.1.046
>>>>> Intel MPI 3.2.0.011
>>>>>
>>>>> current:FOPT:-FR -mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML 
>>>>> -traceback
>>>>> current:FPOPT:-FR -mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML 
>>>>> -traceback
>>>>> current:LDFLAGS:$(FOPT) 
>>>>> -L/usr/local/intel/Compiler/11.1/046/mkl/lib/em64t -pthread
>>>>> current:DPARALLEL:'-DParallel'
>>>>> current:R_LIBS:-lmkl_lapack -lmkl_intel_lp64 -lmkl_intel_thread 
>>>>> -lmkl_core -openmp -lpthread -lguide
>>>>> current:RP_LIBS:-L/usr/local/intel/Compiler/11.1/046/mkl/lib/em64t 
>>>>> -lmkl_scalapack_lp64 
>>>>> /usr/local/intel/Compiler/11.1/046/mkl/lib/em64t/libmkl_solver_lp64.a 
>>>>> -Wl,--start-group -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core 
>>>>> -lmkl_blacs_intelmpi_lp64 -Wl,--end-group -openmp -lpthread 
>>>>> -L/home/xiew/fftw-2.1.5/lib -lfftw_mpi -lfftw $(R_LIBS)
>>>>> current:MPIRUN:mpirun -np _NP_ -machinefile _HOSTS_ _EXEC_
>>>>>
>>>>> Best regards,
>>>>> Wei Xie
>>>>> Computational Materials Group
>>>>> University of Wisconsin-Madison
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> 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 
>>>> <mailto: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 <mailto: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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20101101/5d6b9b02/attachment.htm>


More information about the Wien mailing list