Re: [Boost-users] [MPI] error in hybrid OpenMP + Boost.MPI application

Subject: Re: [Boost-users] [MPI] error in hybrid OpenMP + Boost.MPI application
From: joel falcou (joel.falcou_at_[hidden])
Date: 2010-06-22 16:52:53

Riccardo Murri wrote:
> One MPI rank is started per compute node; all OpenMP threads may call
> mpi::isend(); only one will do mpi::iprobe()/mpi::recv().
>
> Although the above error is in the mpi::communicator::recv(),
> serializing the mpi::isend() calls apparently solves the issue;
> similarly, the program runs fine if I run it on one node only; with
> some other (smaller) inputs, it runs fine as well.  This leads me to
> think that it is a thread-safety issue with the MPI part.  I have
> checked that the MPI library (OpenMPI 1.4.2) is initialized with
> MPI_Init_threads() and provides the threading level
> MPI_THREAD_MULTIPLE.
>
> So, question: is there a (known) thread-safety issue with Boost.MPI,
> or should I definitely look somewhere else?
>   
As a MPI+openMP user, we take great care to not have multiple omp thread 
to do MPI operation not because of Boost.MPI but of MPI way of handling 
things which are not that thread safe.