Been looking through the options to build tensorflow with MPI support. Everything points to a configuration setting that no longer seems to be in the configure script. Has
