More info on tasklist:
  ===============

  When lounching a POE or loadleveler job, with multiple executables
  sharing the same MPI_COMM_WORLD, the "tasklist" determines  
  which component execitables to which processors or nodes.

  A normal tasklist will look somewhat like:

     pop
     pop
     pop
     pop
     pop
     pop
     ccm
     ccm
     cpl

  By specifying "#@ node=6" and "#@ total_tasks=9" in the script, it
  will allocate first 3 nodes (2 procs each) for pop, next 2 nodes
  (1 procs each) and the last node (1 proc) for cpl on NERSC IBM SP
  machine (2 procs per node).  This gives room for ccm and cpl use
  mixed MPI/openMP approach.

  The sample "tasklist" file provided here shows that the model
  executables don't have to be in the same order of component
  names, nor to be continuous, in order to use the MPH module.