[MPICH] debug message?

[William Gropp]

Yes, this is to be expected.  --enable-g=all enables all sorts of  
internal debugging checks, including this memory leak detector.

Bill

On Jan 22, 2008, at 12:41 PM, Wei-keng Liao wrote:

> I got some messages from standard output when my test program called
> MPI_Comm_split(), without MPI_Comm_free() to free the newly created
> communicator. The code that reproduces the message is provided  
> below. The
> messages are
>
> % mpiexec -l -n 4 a.out
> 0: [0] 32 at [0x00000000006daa78], mpid_vc.c[62]
> 1: [1] 32 at [0x00000000006dc3b8], mpid_vc.c[62]
> 2: [2] 32 at [0x00000000006dd648], mpid_vc.c[62]
> 3: [3] 32 at [0x00000000006dd648], mpid_vc.c[62]
>
> No such messages if MPI_Comm_free() is called, though.
>
> I also found that this only happens when I configured  
> MPICH2-1.0.6p1 with
> "--enable-g=all --enable-debuginfo".
>
> Some other messages also appear for different calls, like a data  
> type is
> not freed by the program.
>
> Wei-keng
>
>
> ----   test code in C ----
> #include <stdio.h>
> #include <mpi.h>
> int main(int argc, char **argv) {
>     int rank, np, new_np, color;
>     MPI_Comm new_comm;
>
>     MPI_Init(&argc, &argv);
>     MPI_Comm_rank(MPI_COMM_WORLD, &rank);
>     MPI_Comm_size(MPI_COMM_WORLD, &np);
>
>     color = 1;
>     if (rank < 2) color = 0;
>     MPI_Comm_split(MPI_COMM_WORLD, color, rank, &new_comm);
>
>     /* MPI_Comm_free(&new_comm); */
>
>     MPI_Finalize();
>     return 0;
> }
>

William Gropp
Paul and Cynthia Saylor Professor of Computer Science
University of Illinois Urbana-Champaign


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20080122/75e26cc1/attachment.htm>