Vectorization of Loops with Matrix Multiplications

Hi all. I have the following code which is being repeated many times with different time series and this is producing a huge bottleneck on my code:

%% Example Data
t = 252;
h = randn(252,2);
l = round(0.1 * t,0);

%% Loop
for i = 1:(l - 1)
    o_tmp = h(1,:).' * h(1+i,:);

    for j = 2:(t - i)
        o_tmp = o_tmp + h(j,:).' * h(j+i,:);
    end

    o_tmp = o_tmp / (t - i);
end

I would like ti know if there is a suitable way to improve it, maybe by vectorizing all computations or just some of them. Any suggestion is more than welcome. Thanks in advance for your help!

[David Hill]

It would help if you explained what you are trying to do (big picture).

Answer 1

[Matt Shellhammer]

%% Example Data
t = 252;
h = randn(252,2);
l = round(0.1 * t,0);
%% Loop
for i = 1:(l - 1)
    o_tmp = (h(1:(t-i),:).' * h((1+i):t,:))/(t-i);
end