Hi...
Thank you for replying.
The reason as to why SIGIO might be an issue is that, in the documentation for mpiJava they say that it is just a random signal that they proposed as an alternative for SIGUSR2. Because of the threading nature of Java, this signal has conflicts with Gui's etc. Also they LD_PRELOAD the libjsig to catch signals. The thing is I've sorted out the segmentation fault, but now when I run a java app, simple hello world from all the proc's, it only gives me only the head node, i.e. n0.
My question is. Has there been major changes since version 6.5? Because it seems that LAM 6.5 is the only one that works with mpiJava, from what I can tell. I'll post more info later...
Lee
He who has a why to live can bear with almost any how.
Friedrich Nietzsche
________________________________
From: lam-bounces_at_[hidden] on behalf of Brian Barrett
Sent: Sun 2006/07/16 10:34 PM
To: General LAM/MPI mailing list
Subject: Re: LAM: < lam-7.1.2 and mpi Java >
On Jul 16, 2006, at 1:54 PM, Pool, Lee, Mr <14117614_at_[hidden]> wrote:
> I'm having some trouble with getting mpiJava to work with
> lam-7.1.2! I gives me a segmentation error and says that one of the
> nodes did not call mpi_init, which cannot be the case! Is there any
> reason to suspect a problem with the signal SIGIO?
Can you run C MPI applications with your installation of LAM/MPI? If
not, it would be useful to know how those are failing.
The LAM message that one process didn't call MPI_INIT is pretty hard
to argue with -- according to LAM, at least one process didn't get
far enough through MPI_INIT to register itself as an MPI process.
This happens fairly early in MPI_INIT and is usually indicative of a
failure before MPI_INIT. Although it can occasionally indicate a
failure within LAM. We don't normally test with mpiJava, so I can't
say if there are potential issues there. What makes you think there
is a SIGIO issue? You mention it, but don't include any output that
would suggest an issue there.
Brian
--
Brian Barrett
LAM/MPI developer and all around nice guy
Have a LAM/MPI day: http://www.lam-mpi.org/
_______________________________________________
This list is archived at http://www.lam-mpi.org/MailArchives/lam/
|
