[Wien] wien2k in suse 9.2

Andrea Salguero salguero at itp.uni-frankfurt.de
Wed Apr 27 16:29:46 CEST 2005


dear all
I have seen that there are so many people reporting problems about parallel
jobs, this is a problem I think a little bit different, and perhaps I know
what the problem is but I don't know how to solve it, may be some of the
people that use suse 9,2 or 9.1 can help me. The problem is the following:
I have two machines AMD 64 bits, with suse 9.2, intel compilers for fortran
and c++ , mkl and so on instaled under the same path in both machines, they
are sharing the
same nfs. I have compiled WIEN2K_05 with 64 bits compilers
with the following options:

current:FOPT:-FR -mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML
current:FPOPT:-FR -mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML
current:LDFLAGS:-L/opt/intel/mkl72/lib/em64t -Vaxlib -static-libcxa -pthread
current:DPARALLEL:'-DParallel'
current:R_LIBS:-lmkl_lapack /opt/lib/libgoto.so ../xerbla.o -lmkl_em64t
-lguide
current:RP_LIBS:-L /usr/local/SCALAPACK -L /usr/local/BLACS/LIB -lpblas
-lredist -ltools -lscalapack -lfblac
s -lblacs -lmpi

I can go from one machine into the other with ssh without password and when I
do (from machine1 to machine2):
ssh machine1 date
I obtain the date of the machine2 and viceversa, without password.
I have a .machines file like the following:
1:machine1
1:machine2
granularity:1

and this file works when I run calculations in 32-bits machines with debian.
When I try to run a parallel calculation (only parallelizing k, nothing with
MPI) I find that lapw0, and lapw1 runs in both machines but lapw2 doesn't,
and when I make 'top' in both machines (suppose I send the job from
machine1), in machine2 there's nothing and in machine1 is running lapw1para
but never ends!!!!!!!!, when I kill  lapw1para, lapw2 begins to run in both
machines and again when lapw2 finish stay in lapw2para but never ends
neither.

I have changed in the lapw*para scripts remote shell,
nice and even WIENROOT, but I think there is some problem of comunication
between the machines, does some of the people that is working with suse
machines have some idea????, I really thanks any comment about this.

thanks
andrea

--- 
Andrea Salguero
salguero at itp.uni-frankfurt.de
Johann Wolfgang Goethe-Universität
Institut für Theoretische Physik
Max-von-Laue-Strasse 1
60438 Frankfurt/Main
Germany




More information about the Wien mailing list