[CPMD-list] process in local group is dead..help

[Axel Kohlmeyer]

On Fri, 22 Sep 2006 okan.esenturk at nist.gov wrote:

OS> Hi all,
OS>  I am new in this  and not sure it is a network problem or a setting problem.

please _always_ state what version of CPMD you are using and
what platform you are running on. 

this looks like a 'program is crashing because of input' problem.

for example:
- there is no reason to use the 'costly' gauss-hermite integration for
  elements like you have it in your system. kleinman-bylander should 
  work fine. but this needs to be flagged _only_ for norm-conserving 
  pseudopotentials.
- a plane wave cutoff of 120ry conservatively large for this kind 
  of system with goedecker pseudopotentials. with ultrasoft-psps it
  is plain overkill and it is quite possible that you hit a bug.
  a cutoff of 30ry should be good (or even 25ry, if you crank up
  the density cutoff via setting dual to 5 or 6).
  

OS> When I request a run for same input file for two different ppotentials, one 
OS> (SG_BLYP) runs and the other (VDB_BLYP) fails. 


OS> 
OS> Here is the CPMD output for the failed one..
OS> 
OS>  --initial part then
OS>  ***     DETSP| SIZE OF THE PROGRAM IS    2640/  77644 kBYTES ***
OS> MPI_Recv: process in local group is dead (rank 1, SSI:coll:smp:local comm 
OS> for CID 0)
OS> Rank (3, MPI_COMM_WORLD): Call stack within LAM:
OS> Rank (3, MPI_COMM_WORLD):  - MPI_Recv()
OS> Rank (3, MPI_COMM_WORLD):  - MPI_Bcast()
OS> Rank (3, MPI_COMM_WORLD):  - MPI_Bcast()
OS> Rank (3, MPI_COMM_WORLD):  - main()
OS> 
OS> ----
OS> I have read in one of the list e-mail that this could be due to network talk 
OS> failure. However, it worls for one of the potentials and not for the other one. 

the best way to debug this is to set the cutoff to a ridiculously low
value and try running it serially. this way, you'd see an error message,
that you don't get right now.

cheers,
   axel.

OS> 
OS> Could any one tell me how can I make this thing work?
OS> 
OS> Here is the input file that did not work.....
OS> ---------
OS> # L-Serine unit cell: coordinates created using "mercury".
OS> &CPMD
OS>   GEOMETRY OPTIMIZATION XYZ
OS>   CONVERGENCE INITIAL
OS>    1.0d-5
OS>   CONVERGENCE GEOMETRY
OS>    1.0d-6
OS>   CONVERGENCE ORBITALS
OS>    1.0d-7
OS>   CONVERGENCE ADAPT
OS>    0.02
OS>   CONVERGENCE ENERGY
OS>    0.05
OS> 
OS>   SPLINE POINTS
OS>    3000
OS> 
OS>   STRUCTURE BONDS
OS>   STRUCTURE ANGLES
OS> &END
OS> 
OS> &DFT
OS>   FUNCTIONAL PBE
OS>   GC-CUTOFF
OS>    1.0d-6
OS> &END
OS> 
OS> # Serine unit cell is Orthorhombic P212121,
OS> # so we can use a b c alpha beta gamma:
OS> # CRYST1   8.5710   9.3250   5.6150  90.00  90.00  90.00   P212121
OS> &SYSTEM
OS>   POINT GROUP DELTA=1.0E-6	
OS>    AUTO
OS>   ANGSTROM
OS>   SYMMETRY
OS>    ORTHORHOMBIC
OS>   CELL ABSOLUTE DEGREE
OS>    8.571 9.325 5.615 90.00 90.00 90.00
OS>   CUTOFF SPHERICAL
OS>     120.0
OS> &END
OS> 
OS> &ATOMS
OS> *C_VDB_BLYP.psp GAUSS-HERMITE=50 FORMATTED
OS> LMAX=P
OS>  12
OS>     Coordinates omitted..
OS> *N_VDB_BLYP.psp GAUSS-HERMITE=50 FORMATTED
OS> LMAX=P
OS>  4
OS>     Coordinates omitted..
OS> *O_VDB_BLYP.psp GAUSS-HERMITE=50 FORMATTED
OS> LMAX=P
OS>  12
OS>         Coordinates omitted..
OS> *H_VDB_BLYP.psp GAUSS-HERMITE=50 FORMATTED
OS> LMAX=S
OS>  28
OS>         Coordinates omitted..
OS> &END
OS> 
OS> 
OS> Thanks,
OS> 
OS> Okan Esenturk
OS> 
OS> ------- End of forwarded message -------
OS> 
OS> 
OS> NIST, Optical Technology Division
OS> 301 975 825
OS> okan.esenturk at nist.gov
OS> 
OS> 
OS> _______________________________________________
OS> CPMD-list mailing list
OS> CPMD-list at cpmd.org
OS> http://cpmd.org/mailman/listinfo/cpmd-list
OS> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.