Parallel Programming with MATLAB
John Burkardt
Department of Scientific Computing
Florida State University
28 September 2011

Most laptop and desktop machines now include multiple cores.
Computer clusters are available in which hundreds or thousands
of individual computers are linked together so that certain
kinds of programs can access all the memory and all the cores
for one computation.

MATLAB includes a Parallel Computing Toolbox (PCT) that allows a user
to take advantage of the new parallel computing hardware that
is becoming common.  In the simplest case, the user can simply
run programs interactively on the desktop, but get a 4 or 8-fold
speedup if the program is suitable for parallel execution.

MATLAB also includes a Distribured Computing Server (DCS), which
runs on a cluster.  This enables a MATLAB user to run a parallel
MATLAB program on the cluster, taking advantage of tens or hundreds
of cores at one time.  

The PCS and DCS can cooperate, so that the user can work directly 
from the desktop, and issue commands that are sent to the remote
cluster for execution, with results returned to the desktop when
completed.

In this talk, I will discuss the PARFOR command, which is able
to cause certain simple FOR loops to execute in parallel, and
the more powerful SPMD command, which allows a user to divide
a computation up between a master process and many workers.

Some small examples of such MATLAB programs will be displayed,
and I will also demonstrate how a desktop user can communicate
with a remote cluster, using the system employed by the Virginia
Tech "ithaca" cluster as an example.