This is not necessary stated in the standard or guaranteed by it, it's just my intuition...

I expect that MPI_User_function is invoked by the thread that invokes the MPI function that uses it, and that you need a mutex to touch shared state associated with MPI_User_function when running with MPI_THREAD_MULTIPLE, but not otherwise.

I would hope that MPI implementations do not serialize calls to MPI_User_function in MPI_THREAD_MULTIPLE mode, because there are many cases where MPI_User_function is trivially reentrant and I would like simultaneous progress if I call MPI_Allreduce concurrently from 2+ threads (with different communicators, of course) with the same MPI_User_function.

Jeff

On Thu, May 25, 2017 at 9:35 AM, Phil Ruffwind <rf@rufflewind.com> wrote:

Hello all,

I have another question about MPI_User_function. Does MPI make any
guarantees regarding (a) whether multiple invocations of
MPI_User_function could occur simultaneously within a given process for
a given MPI_Op, and regarding (b) which thread MPI_User_function gets
called on?

The answer to (a) affects whether MPI_User_function would need any
synchronization (like mutexes or semaphores) if it mutates global state.
The answer to (b) affects whether it is possible to use thread-local
state inside MPI_User_function.

I would appreciate any clarifications. Thanks.
Phil
_______________________________________________
mpi-forum mailing list
mpi-forum@lists.mpi-forum.org
https://lists.mpi-forum.org/mailman/listinfo/mpi-forum

Jeff Hammond
jeff.science@gmail.com
http://jeffhammond.github.io/