why mpi deliberate harder common memory erlang deliberate easier, both message-passing?

there's lot seductiveness days erlang libel minute together programs multicore. i've listened remonstrate erlang's message-passing denote easier way widespread shared-memory models such threads.

conversely, high-performance computing village widespread together programming denote mpi, also implements message-passing model. nonetheless hpc world, message-passing denote generally deliberate unequivocally challenging way in, remonstrate common memory models such openmp upc easier way in.

does anybody know since there such inadequacy notice message-passing vs. common memory hpc worlds? due component inadequacy erlang mpi exercise summary flitting creates erlang-style message-passing many easier mpi? there reason?