Is there an example on how to use parallel programming with Parallel Computing Toolbox to do a simple matrix computation?

I would like to know how to compute A*B using the Parallel Computing Toolbox 4.0 (R2008b).

Answer 1

[MathWorks Support Team]

This example computes A*B over four workers using the local scheduler. Note that in this simple example the communication outweighs the actual computation, so it is not a useful speed benchmark.

The code for the task is as follows

function out = multtrans(x,y)

% Codistribute the input matrices

A = codistributed(x,'convert');

B = codistributed(y,'convert');

% Multiply local parts

C = A * B;

% Gather the result on lab 1

out = gather(C,1);

end

The code that is run on the client is

% Declare the two matrices on the client. The matrix must preserve innner product convention: 'm x n' matrix can multiply only with 'n x p'

X = [1 2;3 4;5 6];

Y = [7 8 9;10 11 12];

% Pass sections to workers

jm = findresource('scheduler', 'type', 'local');

pjob = createParallelJob(jm);

tsk = createTask(pjob, @multtrans, 1, {X,Y});

pjob.MaximumNumberOfWorkers=4;

pjob.MinimumNumberOfWorkers=4;

set(pjob, 'FileDependencies', {'multtrans.m'});

submit(pjob);

waitForState(pjob);

result = getAllOutputArguments(pjob);

destroy(pjob);

To access the result, execute the following on the client

result{1}