Hello,

 

I have been experimenting with mpich’s implementation of one-sided communication with no remote synchronization that was introduced in the mpi3.0 spec and I have run into a hang issue during connect/accept MPI scenarios.

 

Here is the situation,  I have two applications that connect to each other after they have been launched separately with mpiexec on two different hosts.

 

After connecting, both hosts allocate memory and run MPI_Win_Create as the following:

Host1_program:

   void* buffer = mmap(0, buf_length, PROT_READ | PROT_WRITE, MAP_SHARED|MAP_POPULATE|MAP_ANON, -1, 0);

    MPI_Intercomm_merge(connected_comm, 1, &comm);

    MPI_Comm_rank(comm, &myrank);

    MPI_Win_create(NULL, 0, 1, MPI_INFO_NULL, comm, &win);

Host2_program:

   void* buffer = mmap(0, buf_length, PROT_READ | PROT_WRITE, MAP_SHARED|MAP_POPULATE|MAP_ANON, -1, 0);

    MPI_Intercomm_merge(connected_comm, 1, &comm);

    MPI_Comm_rank(comm, &myrank);

    MPI_Win_create(buffer, buf_length, 1, MPI_INFO_NULL, comm, &win);

 

 

Once the MPI_Window has been created I execute the following on host1:

 Host1_program(continued):

       err = MPI_Win_lock(MPI_LOCK_SHARED, target_rank, 0, win);

        if(err != MPI_SUCCESS)

        {

            return 1;

        }

        err = MPI_Put(buffer, buf_length, MPI_BYTE, target_rank, 0, buf_length, MPI_BYTE, win);

        if(err != MPI_SUCCESS)

        {

            return 1;

        }

        err = MPI_Win_flush(target_rank, win);

        if(err != MPI_SUCCESS)

        {

            return 1;

        }

        err = MPI_Win_unlock(target_rank, win);

        if(err != MPI_SUCCESS)

        {

            return 1;

        }

While this is running, host2’s program is simply sleeping in a getchar().

 

This should run and complete, however, if the other host has not called another mpi function such as MPI_Recv and instead is in a sleep, getchar(), etc. the MPI_Win_flush above will simply hang.

 

Why is the one-sided operation hanging until another MPI operation is called on the target host? My understanding is that utilizing MPI_Win_Flush does not require any action by the target host to complete the operation.

 

This is a stack trace of where the flush gets stuck:

#0  0x0000003b456dea28 in poll () from /lib64/libc.so.6

#1  0x00007fe1c08004ee in MPID_nem_tcp_connpoll ()

   from /usr/local/lib/libmpich.so.12

#2  0x00007fe1c07efab7 in MPIDI_CH3I_Progress ()

   from /usr/local/lib/libmpich.so.12

#3  0x00007fe1c07d35ff in MPIDI_CH3I_Wait_for_lock_granted ()

   from /usr/local/lib/libmpich.so.12

#4  0x00007fe1c07d88bf in MPIDI_Win_flush () from /usr/local/lib/libmpich.so.12

#5  0x00007fe1c08c1b46 in PMPI_Win_flush () from /usr/local/lib/libmpich.so.12

#6  0x000000000040115c in main ()

 

According to this it seems that somehow the remote host is holding something after MPI_Win_Create that in my scenario gets unlocked if I call MPI_Recv on the connected_comm from above after the MPI_Win_Create.

 

Thank you for your time and I look forward to your reply.

 

Respectfully,

Neil Spruit

P.S. MPICH version info

    Version:                                 3.1

    Release Date:                        Thu Feb 20 11:41:13 CST 2014

 

_______________________________________________
discuss mailing list     discuss@mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss