parfor is not really speeding up the function.

I am trying to speed up my code that is computing the max. coherence of each pair of rows insede of matrices A and B:

A = rand(1000, 100);
B = rand(4, 100);


fcmat = zeros(size(A,1), size(B,1));
for i = 1:size(A,1)
    for j = 1:size(B,1)
        c = mscohere(A(i,:),B(j,:));
        fcmat(i,j) = max(c);
    end
end

I have tried to use a simple parfor, but it runs in approximately same time as the non-parfor loop

B_n = size(B,1);
parfor (i = 1:size(A,1))
    for j = 1:B_n
        c = mscohere(A(i,:),B(j,:));
        fcmat(i,j) = max(c);
    end
end

Is there anything I have done (or probably have not done) that is causing this code to not run faster? Is there some other way/technique that could make this code run faster?

Thank you.

Answer 1

[Ganesh Gudipati]

Hi Filip,

I could see some significant speed up in the code you provided after using parfor. Here are my observations

However you can check the following documentation to Improve parfor Performance. I hope this helps.

Thanks

Ganesh Gudipati

[Filip]

Thank you.

I have found the cause of the problem at https://www.mathworks.com/matlabcentral/answers/13507-does-parfor-make-a-difference-without-parallel-computing-toolbox

The problem was that Matlab allows the user to use parfor without having the Parallel Computing toolbox installed.