On Feb 8, 2005, at 8:17 AM, Eiso AB wrote:
> /software/bin/cyana2Jonathan Herriott wrote:
>> Scratch that. Apparently the script that was written to run this
>> benchmark (I didn't write it) was used under csh, which seems to have
>> a problem with redirecting a file as input to an MPI application. I
>> don't know exactly why, but when I rewrote the script in sh, it is
>> actually running.
> unfortunately it's already in sh
>
> I was wondering, is it necessary to specify at lam/mpirun compile time
> if you're going to use rsh or ssh?
You can specify the default, but you can override that at run-time with
the LAMRSH environment variable.
Keep in mind that LAM's mpirun does not use rsh or ssh -- it uses the
LAM daemons. The LAM daemons are initially setup with rsh, ssh, or one
of several other remote agent technologies (e.g., bproc, SLURM, PBS).
So if you're able to lamboot, then you rsh/ssh issues are working
properly.
stdin is a different issue altogether, and unfortunately, at least
partially out fault. :-(
We accidentally broke stdin forwarding in v7.1. It is not fixed in
7.1.1, but is fixed in subversion and our nightly tarballs (i.e., what
will shortly become 7.1.2). You can download and try that:
http://www.lam-mpi.org/svn/ -- this tarball is what will shortly become
7.1.2.
We mainly only have doc fixes left for 7.1.2, but we were testing the
newest version of Libtool (1.5.12) and found a problem with how it
links threading libraries (see
http://lists.gnu.org/archive/html/libtool/2005-02/msg00058.html).
There will be a new release of Libtool (v1.5.14), but it won't happen
until this upcoming weekend at the earliest, meaning that it'll take us
at least a little more time (assuming everything else goes well) before
we can release 7.1.2.
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
