[Pw_forum] error in the q2r.out

Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu
Tue Sep 25 13:37:02 CEST 2007

On Tue, 25 Sep 2007, [gb2312] ´÷¼ÑîÚ wrote:

JD> >cheers,
JD> >  axel.
JD> I have the same problem when i run mpirun -np 4 ph.x -npool 2 < *.ph.in . But if i

please when running with mpirun you should use the -in flag instead
of redirecting input from stdin (and using filename globbing, i.e. 
filename expansion with *, is even more problematic as it may redirect
multiple files). not all MPI packages replicate stdin to all nodes.

JD> run mpirun -np 4 ph.x -npool 1 < *.ph.in  or mpirun -np 4 ph.x < *.ph.in, the
JD> error would disappear. Is it the meaning of the memory error? what did you mean

do generate a CRASH file? have you looked at it? it probably means, 
that the calculation that you were asking for, does not support 
parallelization over k-point pools.

 JD> "in serial"?

"in serial" means "not parallel"

there is no memory error. when your remote processes die,
your local process cannot write to their corresponding
sockets anymore. this is one of the (many) drawbacks of
using MPICH: the error reporting of this MPI packages
is not very good and frequently it "eats" the last lines
of output of the crashing application (which would be 
essential to understand what is going on).


JD> Thanks,
JD> Jiayu
JD> ------------------------------
JD> Jiayu Dai
JD> National University of Defense Technology, P R China

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.

More information about the users mailing list