hi all... once again in need of help...
as a normal user on the system, i have compiled AMBER 6 with the machine
file Machine.g77_mpich with no changes. i installed it as a normal user in
my user directory in order to make sure that all the permissions were user
level, then as root i moved everything to /usr/local/amber6 (=$AMBERHOME
and is also a NFS across the cluster, and accessible to all users...
hopefully).
i have copied some of the test files $AMBERHOME/test/4096wat/
to a working directory under $HOME.
when i run the benchmark file, with $DO_PARALLEL='mpirun -np 8 -nolocal',
i get the following output:
in $HOME/Data
prompt> ./benchmark
write start: can't write file
apparent state: unit 6 named bench.out
last format: (/,2x,a,i4,a,a)
lately writing sequential formatted external IO
Abort
/bin/rm: cannot remove `restrt': No such file or directory
i am assuming that the write start problem is either a gbin issue or the
output file bench.out cannot be written... seems the latter as it is not
present after the program aborts. confusing as i have write privileges in
this directory, all the files copied are mine, and the account i am using
has rsh access across the entire cluster... ? i understand that the file
`restrt' is not present to be removed.
i don't understand why the program keeps aborting.
i have another account which does not have rsh access across the cluster,
but is instead an NFS partition across the cluster... and i have done the
same thing from this account with the same output. i have run the copy of
the benchmark script in the NFS account from the account which has rsh
cluster privilege with the same result.
i am fairly sure that MPICH is installed correctly, but somehow getting it
and AMBER to communicate is problematic.
any and all help for a confused muggle greatly appreciated,
charles
Received on Tue Nov 13 2001 - 09:26:26 PST