We could have used exact knowledge of the amount of data on each process (the
lcnt value in the solution) instead of using MPI_Gather to
collect the values. Using MPI_Gather, however, is simple and allows us to
ensure that we are using the correct values.