same poll barrier implementation as in the C++ layer in triqs. Reduces CPU load on waiting MPI ranks drastically. Default wait time is 1 msec as in the C++ implementation. passing 0.0 or None to the barrier function call will use the default mpi.barier() call of mpi4py. I added a very brief test in one of the few mpi enabled python tests to check the function and cleaned up a few white spaces.
same poll barrier implementation as in the C++ layer in triqs. Reduces CPU load on waiting MPI ranks drastically. Default wait time is 1 msec as in the C++ implementation. passing 0.0 or
None
to the barrier function call will use the default mpi.barier() call of mpi4py. I added a very brief test in one of the few mpi enabled python tests to check the function and cleaned up a few white spaces.