[Wien] MPI

[Laurence Marks]

Hmmm.

This is my parallel_options on a machine with openmpi:

setenv USE_REMOTE 1
setenv MPI_REMOTE 0
setenv WIEN_GRANULARITY 1
setenv WIEN_MPIRUN "mpirun -x LD_LIBRARY_PATH -x PATH -np _NP_
-machinefile _HOSTS_ _EXEC_"
set a=`grep -e "1:" .machines | grep -v lapw0 | head -1 | cut -f 3 -d:
| cut -c 1-2`
#echo $a
setenv MKL_NUM_THREADS $a
setenv OMP_NUM_THREADS $a
setenv MKL_DYNAMIC FALSE
#echo MKL $MKL_NUM_THREADS $OMP_NUM_THREADS $MKL_DYNAMIC
if (-e local_options ) source local_options
set remote = "/bin/csh $WIENROOT/pbsh"

The command pbsh should be in SRC_mpiutil at
http://www.wien2k.at/reg_user/unsupported/ which has a few other
things, let me know if I forgot to include it. It is

mpirun -x LD_LIBRARY_PATH -x PATH -np 1 --host $1 /bin/csh -c " $2 "

What this does is use mpirun locally for everything rather than
communicating via ssh and spawning processes which can sometimes be a
problem with openmpi.


On Wed, Feb 20, 2013 at 10:45 AM, Luis Ogando <lcodacal at gmail.com> wrote:
> Dear Prof. Marks,
>
>    Thank you very much for your prompt answer.
>    I am using openmpi, but I believe that I am facing some of the tricky
> issues you mentioned. I work in a SMP machine and the calculation starts
> fine. After some tens of iterations, MPI suddenly asks for a password and
> everything goes down to the drain.
>    I am using open mpi 1.6. Do you recommend any older version ??
>    All the best,
>                   Luis Ogando
>
>
>
>
> 2013/2/20 Laurence Marks <L-marks at northwestern.edu>
>>
>> One that works.
>>
>> Some versions of openmpi have problems although that is probably the
>> best option for the future. There are some tricky issues with openmpi
>> related to how your flavor of ssh works, there is no standard and some
>> do not propogate kill commands which means that they can leave
>> orphans.
>>
>> An alternative is mvapich. In a benchmark that I did a few months ago
>> the Intel mpi was much better for AVX instructions, but that may have
>> changed.
>>
>> Openmpi is easy to compile; mvapich can be a little trickier.
>>
>> N.B., if you have fast connections, e.g. infiniband, they are more
>> than fast enough and I have never seen this as rate limiting with
>> Wien2k. With ethernet it matters.
>>
>> On Wed, Feb 20, 2013 at 10:29 AM, Luis Ogando <lcodacal at gmail.com> wrote:
>> > Dear Wien2k community,
>> >
>> >    Is there any recommended flavor and version of an MPI compiler to use
>> > with " Intel(R) Fortran Intel(R) 64 Compiler XE for applications running
>> > on
>> > Intel(R) 64, Version 12.0.3.174 Build 20110309 " ?
>> >    All the best,
>> >                       Luis Ogando
>>
>>
>>
>> --
>> 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
>
>



-- 
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