MPI_Allreduce error

[mchatzikos]

This is an edge case that goes beyond the purpose of graspg. Still, it might be better if the reported problem were addressed.

The script attached at the bottom of this message computes the ground state of He-like Zn. Running on 1 processor (NP=1) gives the same result as a traditional grasp run. Running on 2 processors leads to the following MPI error:

[server:256262] *** An error occurred in MPI_Allreduce
[server:256262] *** reported by process [3773693953,0]
[server:256262] *** on communicator MPI_COMM_WORLD
[server:256262] *** MPI_ERR_TRUNCATE: message truncated
[server:256262] *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort, 
[server:256262] ***    and potentially your MPI job)

The bug comes from the following call starting on line 43 of cofpotmpi.f90:

 42 
 43       CALL MPI_Allreduce (YP, tmpmpi, npts, MPI_DOUBLE_PRECISION,  &
 44                           MPI_SUM, MPI_COMM_WORLD, ierr)
 45       CALL dcopy (npts, tmpmpi, 1, YP, 1)

I did not do much debugging after this point. I did place a CALL MPI_Barrier() on line 42, which rank 0 never reached. The run stalls with a bunch of messages like:

mpirun: Forwarding signal 20 to job
mpirun: Forwarding signal 18 to job

I hope this helps a bit.

He-like_Zn.zip

EDIT:
The contents of the isodata file are:

Atomic number:
   30.000000000000000     
Mass number (integer) :
   64.000000000000000     
Fermi distribution parameter a:
  0.52338755531043146     
Fermi distribution parameter c:
   4.4057704883852029     
Mass of nucleus (in amu):
   63.912684602909998     
Nuclear spin (I) (in units of h / 2 pi):
   0.0000000000000000     
Nuclear dipole moment (in nuclear magnetons):
   0.0000000000000000     
Nuclear quadrupole moment (in barns):
   0.0000000000000000