On Feb 15, 2006, at 2:16 PM, Nick Rasmussen wrote:
> I'm running into a situation where lam goes into the spinlock
> in lam_ssi_rpi_usysv_proc_read_env, and spin-waits forever.
> This is repeatable, but small perturbations to the resolution
> of the simulation typically cause this issue to go away.
>
> We're currently using lam-7.1.2b28 (we ran into other bugs in
> the 7.1.1 release version) on suse9.3, x64_64.
>
> Is this a known bug? Is there a later beta available that
> has a fix for this?
There's a b31, but there were not changes to the usysv RPI.
Can you describe the behavior about how this happens? For example,
are you sure that a matching send has been posted? (e.g., the
corresponding send was on the same communicator and sent with the
expected tag)
--
{+} Jeff Squyres
{+} The Open MPI Project
{+} http://www.open-mpi.org/
|
