Problem solved.
The program has a bug. Instead of using the status variable (array) the
program uses the stat variable (scalar). Therefore there is a buffer
overflow.
On Monday 14 August 2006 09:58, Sven Stork wrote:
> The problem is that after the MPI_Recv call the loop
> variable is set to
The problem is that after the MPI_Recv call the loop
variable is set to 0, and therefore the loop doesn't stop
as supposed to be for 2 processes.
If you use another temp. varaible as parameter the program
works. The strange thing is that temp. variable wont be changed and still
contains the or
I'm not quite sure I understand -- does the application hang in an MPI call?
Or is there some compiler error that is causing it to execute a DO loop
incorrectly?
On 8/11/06 6:25 AM, "Sven Stork" wrote:
> The real problem is not the -g it is the -O0 option which will be
> automatically added by
The real problem is not the -g it is the -O0 option which will be
automatically added by -g. If you compile with "-g -ON" for 0 < N everythings
works as expected.
Thanks,
Sven
On Friday 11 August 2006 11:54, Bettina Krammer wrote:
> Hi,
>
> when I use the attached hello.f with Open MPI 1.1.0 a
Hi,
when I use the attached hello.f with Open MPI 1.1.0 and underlying Intel
9.0 or 9.1 compiler on our Xeon cluster, it is deadlocking when compiled
with -g option but works without -g:
===
output with -g:
$mpirun -np 2 ./hello-g
My rank is0 !
waiting for messag
