On Tue, 5 Nov 2002, Ziriax John M Civ NHRC/DET wrote:
> I rebuilt the lam rpms using the FC environment varialble prointing to
> the Leahey Fortran compiler without problems. However when I went to
> install the new RPMS on the slave nodes they complained about the lack
> of libaio.so.1.
Yes, that would be required. Sorry -- didn't think to mention that.
> I did have to use the switcher and opium commands as you thought.
Good. Switcher is your friend. :-)
> /usr/bin/ld: cannot find -laio
>
> And lamtest.out contains:
Ok. So for some reason, even though you installed libaio on the nodes,
it's not being found when MPI programs are being compiled. The mpicc
wrapper compiler is implicitly adding "-laio" to compile lines (do "mpicc
-showme" to see what flags mpicc adds), and for some reason, ld is not
finding it.
I'm guessing that you didn't install the same libaio RPM on the clients as
is installed on the head node.
You can find out which RPM provided libaio.so.1 on the head node by
executing:
rpm -qf /path/to/libaio.so.1
This will give you an RPM name. If it's different than what is installed
on the clients, uninstall the libaio RPM on the clients and install the
same one as is on the head node, and that should solve your problem.
Sidenote: it'll probably be easier to test manually rather than having to
run the entire OSCAR test suite every time. Take a simple MPI program
(e.g., a "hello world" kind of program) and try to compile it with LAM's
mpicc wrapper compiler. When you can compile and run it, the OSCAR test
will most likely pass.
{+} Jeff Squyres
{+} [EMAIL PROTECTED]
{+} http://www.lam-mpi.org/
-------------------------------------------------------
This sf.net email is sponsored by: See the NEW Palm
Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en
_______________________________________________
Oscar-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/oscar-users
