Is element-wise multiplication or element-wise squaring faster?

I have a matrix A for which I need its individual elements squared. Which of the following is faster?

% This... >> B = A .* A;

% Or this? >> B = A .^ 2;

Anyone know which of these is faster? Some initial tests suggest the direct squaring is marginally faster... but I'd like to know why.

Thanks!

Answer 1

[Kelly Kearney]

A slightly more robust timing test, using timeit:

n = 1:100:1000;
t = zeros(length(n),2);
for ii = 1:length(n)
    a = rand(n(ii));
    t(ii,1) = timeit(@() a.*a);
    t(ii,2) = timeit(@() a.^2);
end

plot(n, t, 'marker', '.');
legend('*', '^');

Playing around with the values of n give me different results. For small matrices (n ~ 1-10), times is faster. For larger matrices (n ~ 100-4500), power is faster. After this point, the two converge to nearly identical run times.

Not sure on the why, since both function are built-ins. But running similar tests to the above should give you a good idea of how your computer will respond for the matrix sizes relevant to your problem.