[Wien] problem in parallel caculation

hassan jamnejad hassan.jamnejad at gmail.com
Tue Feb 4 10:12:15 CET 2014


Dear Prof. Blaha,

    Thank you so much for your kind response.I would like to share my
experience with others who may probably have the same problem.

    With the guidelines you mentioned I used the k-point palatalization
(unfortunately fine grained did not worked as you mentioned for more than a
few CPUs) and  I changed the ".machines" file to something like:

<<
    lapw0:localhost:2
    1:localhost
    .
    .
    .
    1:localhost
    granularity:1
>>

  I set the number of hosts to the number proposed by testpara.

  I also set OMP_NUM_THREADS=1 to be sure that each process is running in
just one thread and have added the "-O3 -mfpmath=sse,387 -ffast-math" to
compiler options.Also using the OpenBLAS library.

  with this changes I managed to run each complete spin polarized cycle in
around 15minutes. but when I try to run scf in " -orb -p" option the
following errors appears:

(in dayfile)

**  LAPW2 crashed!
error: command   /root/WIEN/lapw2para -up uplapw2.def   failed


(in STDOUT)

At line 331 of
 file sumpara.f (unit = 10, file = 'dimer.dmatup_1')
Fortran runtime error: Bad integer for item 0 in list input
 >>>>>>>>           1
cp: cannot stat `.in.tmp': No such file or directory

>   stop error


I will be very grateful for guiding me bout my problem


kindly regards

Hassan Jamnejad
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20140204/b656be00/attachment.htm>


More information about the Wien mailing list