Dear Vasilii,
I think that stdin redirection under MPI is not defined in the
standard, and thus each implementation makes its own thing. In my case
(MPICH2 version 1.0.3) they only supported slow, low volume stdin input.
I had to give the input file as an argument instead of redirecting
from stdin adding these lines at the very end of the variable definition
in siesta.F:
character arg1*100
call getarg(1,arg1)
open(5,FILE=arg1)
I can imagine that in your case stdin might be redirected to the
non-master node for whatever reason. In any case, it's easy to give it a
shot...
Xavier
Vasilii Artyukhov wrote:
Hi everyone,
I seem to have a very strange problem with parallel SIESTA. The make
goes okay, but the program does not read the input .fdf file. In fact,
all it does is output the out.fdf file that says:
fdf-debug 0 # Default value
SystemName # Default value
SystemLabel siesta
# Default value
WriteXML T # Default value
alloc_report_level 0 # Default value
xc.functional LDA # Default value
xc.authors PZ # Default value
user-basis F # Default value
user-basis-netcdf F # Default value
PAO.BasisSize standard # Default value
PAO.BasisType split # Default value
Number_of_species 0 # Default value
So it's all default values, which means that the input file has been
completely ignored. Have any of you had the same problem? Could there be
some important details that I missed during the compilation? The system
is an Itanium2 cluster, and the serial version of SIESTA works on it
with no problems at all.
Thanks everyone!