Homework 3
due: Friday, Feb 4

Write a parallel version of the explicit time-dependent heat equation code
that we prototyped in serial in class. 

Include the following:
 - an implementation using both blocking and non-blocking communication
 - a 2d spatial decomposition that minimizes local work/communication
 - ability to run on any number of processors, though as usual
   you can restrict the grid to simplify the bookeeping
 - an i/o strategy that writes output files in the same order
   as the serial code
 - strong and weak scaling studies on fd
 - global calculation of total "energy" sum(T^2) every selected
   number of timesteps
- an input parameter file (rather than argc/argv)
- 2d gaussian initial condition
- graphics of the solution (including code that produced it)

Do not worry about:
  - making your code run for any arbitrary number of procs. It
    is ok to specify that it can only run on even numbers, powers
    of two, etc. But you most document it and do error checking within
    program