On 08/02/2011 06:27 AM, Henrik Löfås wrote:


2011/8/2 Hongyi Zhao <[email protected] <mailto:[email protected]>>


    Any furthermore, if some stuff  are  comments,  I think  a # symbol
    should be put before it.   I know the fdf format itself is very
    flexible,  but I think, even  so,  we should also follow some common
    programming practices ;-)  Otherwise, it maybe so  elusive for us to
    understand the fdf file  and make fdf file looks confusing,
    especially for newbies ;-)
    For example,  on this list, I've seen for several times there are
    people who like to use lines like the following in the fdf files:

    ******************************__********
    or
    ==========================

    But I think this is not a good programming practice though they will
    not  affect the normal operation of siesta ;-)


Yes, this is just a matter of taste.



    I also think so, but I've tried to do so with a even large box but
    failed to run it at the end.
    Say, I've tried with the following box:

    --------
    LatticeConstant 1.0 Ang
    %block LatticeParameters
      35.000000  35.000000  35.000000  90.000000  90.000000  90.000000
    %endblock LatticeParameters
    ------------



    And I meet the following errors:


    -----------
    MINVEC: BASIS VECTORS ARE LINEARLY DEPENDENT
    MINVEC: BASIS VECTORS ARE LINEARLY DEPENDENT
    ERROR STOP from Node:   13

    MINVEC: BASIS VECTORS ARE LINEARLY DEPENDENT
    ERROR STOP from Node:    9
    redata: Net charge of the system         =     0.0000 |e|
    redata: Max. number of SCF Iter          =       50


As for your second error, have you checked your compilation of SIESTA?
For me the files you sent run without any problem.

*second error*?  which do you refer to?

Do you mean the one with the following setting:

-----------
LatticeConstant 1.0 Ang
%block LatticeParameters
  35.000000  35.000000  35.000000  90.000000  90.000000  90.000000
%endblock LatticeParameters
--------------



As it seems the nodes crashes independently which I guess indicates that
it is something wrong with the parallell-compilation. Have you tried to
run it as a serial job?

I've run both of them with the NP=1, and get the same errors.

Any hints?

Regards

Best regards
Henrik




--
Hongyi Zhao <[email protected]>
Institute of Semiconductors, Chinese Academy of Sciences
GnuPG DSA: 0xD108493

Responder a