By skimming through the code, this seems to have the same data-flow model as Intel's TBB, but extended to distributed memory configurations using MPI as backend.