The ABINIT Users Mailing List ( CLOSED )

[Torsten Hoefler <torsten.hoefler@cs.tu-chemnitz.de>]

Hi,
> -P-0000  leave_test : synchronization done...
> kpgio: loop on k-points done in parallel
>  newkpt : in paral mode, only node 0 describes in the main output file the 
> copying of wfs.
> -P-0000  wfconv:   235 bands initialized randomly with npw= 11857, for 
> ikpt=     1
> -P-0000  wfp12_3203:  p4_error: net_recv read:  probable EOF on socket: 1
> rm_l_12_3204: (3372.113281) net_send: could not write to fd=5, errno = 32
> conv:   235 bands initialized randomly with npw= 23727, for ikpt=     2
> -P-0000  wfconv:   235 bands initialized randomly with npw= 23749, for 
> ikpt=     3
> p0_29430:  p4_error: interrupt SIGx: 13
this is an MPICH/MPICH2 error in MPICH's p4 (TCP) device. This occurs
clearly inside the MPI Layer (but may be caused from upper layers by
passing a wrong memory pointer so MPI). But it looks like your TCP 
breaks down/misbehaves. Which operating system are you using?

I would recommend to try another MPI implementation if possible to see
if the error is code or MPI related. You could try:
   1. Open MPI (www.open-mpi.org)
   2. LAM/MPI (www.lam-mpi.org)

Best,
  Torsten

-- 
Torsten Hoefler                                  bash$ :(){ :|:&};: 
address: Indiana University    | http://www.indiana.edu
         Open Systems Lab      | http://osl.iu.edu
         150 S. Woodlawn Ave.  | Bloomington, IN, 474045-7104 | USA
         Lindley Hall Room 415 | +01 (812) 856-1820