On Mar 18, 2007, at 8:30 PM, wkbrowne_at_[hidden] wrote:
> Hello all,
> I have a very basic and perhaps naive question regarding running a
> lam/mpi
> program. I am compiling code that I plan to use as an end user.
> The code
> compiles perfectly using mpicc, but when I run it it reports the
> following:
> MPI process rank 0 (n0, p18977) caught a SIGSEGV in
> MPI_Errhandler_set.
> Rank (0, MPI_COMM_WORLD): Call stack within LAM:
> Rank (0, MPI_COMM_WORLD): - MPI_Errhandler_set()
> Rank (0, MPI_COMM_WORLD): - main()
> I've compiled the code on another machine using lam-mpi and it
> works just
> fine, so I strongly suspect the issue is my usage of mpi. My only
> guess
> is that this is a memory allocation problem. Any suggestions would be
> greatly appreciated.
You're probably correct that it's a problem with memory allocation.
I'm not exactly sure why it would segfault there (I'd almost expect
it to do so when creating the error handler). Can you compile LAM
with the --with-purify configure option and try running your
application under valgrind? Would probably help to find the memory
error.
Brian
|
