Low computation efficiency of simple sentence in MATLAB

Hi, everyone

I have an computation efficiency problem here, I wish someone could help me out.

here is part of my MATLAB code (the actual code is much bigger than the following):

lambda_new1=ones(32*32*32,1);

W1=zeros(32*32*32,t);

W2=zeros(32*32*32,t);

W3=zeros(32*32*32,t);

W4=zeros(32*32*32,t);

temp=reshape(sum(repmat(lambda_old1,1,4).*[W1(:,i),W3(:,i),W2(:,i),W4(:,i)]),2,2);

How come that the last sentence in above code is running so slow, which accounts for 35% total running time? Actually my code is much bigger than the above and there are more complicated computation, but why such a short and concise computation takes so much time?

Is there any better way to improve computation efficiency?

Also the line numbers in my MATLAB (2016a and 2016b) would disappear after a while, if I relaunch the software, the line number would come back. Is this an intrinsic bug in MATLAB? How can I fix it?

Greatly appreciate your help

Thank you

Albert Liew

Answer 1

[Thorsten]

Instead of

 X = repmat(lambda_old1,1,4).*[W1(:,i),W3(:,i),W2(:,i),W4(:,i)];

if might be faster to use

 X = bsxfun(@times, [W1(:,i),W3(:,i),W2(:,i),W4(:,i)], lambda_old1);

[Albert]

Hi, Thorsten

Thank you very much for your answer. It does work much faster. The reason why is that, I guess, by using 'bsxfun', we simply circumvent the process of copying and pasting a large matrix, which is time consuming.

Thank you again.

[Thorsten]

Exactly.