triqs.utility.mpi_mpi4py.barrier

triqs.utility.mpi_mpi4py.barrier(poll_msec=1)

Use asynchronous synchronization, otherwise mpi.barrier uses up all the CPU time during the run of subprocess. if poll_msec is 0.0 or None the default mpi barrier of mpi4py is used. :param poll_msec: :type poll_msec: float, time step for pinging the status of the sleeping ranks in msec