The ABINIT Users Mailing List ( CLOSED )

[Alain Jacques <ajs.jacques@gmail.com>]

Hmmm ... quite specific. You mentionned MVAPICH ... then I assume MPI
level 1. So you may try to add the configure option --with-mpi-level=1
... but it definitely stinks :-) Tell us if it does some good.
And did you try one of the abinit parallel tests (if no, copy the
t_kpt+spin.in, t_kpt+spin.files and 26fe.pspnc files to a directory,
modify the .files and test abinip with them)? Do you experience the
crash only with your own input file?


Kind regards,

Alain

BTW ... download the 5.8.4p release, it solves a nasty bug (unrelated to
mpi)


deyulu@yahoo.com wrote:
> Dear all:
>         I have some trouble to run the parallel job with abinit 5.8.4. The
> machine has 
> Two Dual-Core Xeon 5100 processors per node, and I compiled the code with 
> intel
>
> compilers and MPI MVAPICH (mvapich) with the following parameters in config:
>
>   $ ./configure --disable-all-plugins --enable-mpi=yes --enable-gw-dpc
> FC=mpif90 CC=mpicc CXX=mpiCC FCFLAGS=-O2 -assume byterecl -xT CFLAGS=-O2 -xT
>
>         The tests with abinis seem to be fine,  so do the ones with abinip
> using 1 node
> (4 cpus).  For jobs using 2 nodes (8 cpus),  most of the time the program
> couldn't 
> read properly the input file with only a few lines in the log file:'
>
> "  ABINIT
>
>   Give name for formatted input file:"
>
>        The err file contains the following messages:
>
> "lslogin1% more err.o281765
> cd /work/teragrid/tg455780/silane.abinit
> ibrun /work/teragrid/tg455780/abinit-5.8.4/src/98_main/abinip
> forrtl: severe (24): end-of-file during read, unit 5, file stdin
> Image            PC                Routine            Line        Source
>
> abinip                   00000000013B79FE  Unknown               Unknown  
> Unknown
> abinip                   00000000013B6A58  Unknown               Unknown  
> Unknown
> abinip                   000000000135EB86  Unknown               Unknown  
> Unknown"
>
>         I understand that this may be specific to the machine, especially 
> the
> MPI environment, and I'm also contacting with the technical staff. I'm
> wondering if 
> someone has experienced similar problems and may have a solution for it.
>
>
>        Best regards,
>        Deyu Lu
> -------------------------------------------------------------------------------------------------------
> Deyu Lu (Ph.D)
>
> 190 Chemistry Building
> University of California, Davis 
> One Shields Avenue
> Davis, CA 95616
> Group Webpage: http://angstrom.ucdavis.edu/
> ---------------------------------------------------------------------------------------------------------
>