We introduced various kinds of parallel computers
and networks.
We are focusing on SPMD (single program, multiple
data).  MPI is a library that enables several
machines to run different instances of the same program,
where those instances talk to each other.

We did two example codes in Python that use the midpoint rule of 
integration.
The first is mid1.py and is very naive.  The communication time is
proportional to the number of processors, regardless of the
network topology.  However, it relies only on basic send and recv
messages.

The second, mid2.py, introduces collective communications "broadcast" 
and "reduce".  These higher-level calls allow the library to exploit 
network topology to optimize the time it takes to do the entire 
communication.

The file hosts is used to tell mpirun what machines are available for 
logging on to via the -machinefile <file_name> option.