Making a code runs faster

How can I make the following code runs faster (This a part of a larger code). I do not want to use parfor.

    for rotation=1:360
        RHO(:,1)=(CL*Landa)./T(tnum).LoAn(:,1);
        THETA=T(tnum).LoAn(:,3);
        X(:,1)=RHO.*cos(deg2rad(THETA(:,1)+rotation));
        Y(:,1)=RHO.*sin(deg2rad(THETA(:,1)+rotation));
        XP=round(72+X);
        YP=round(72+Y);
        mat=zeros(144,144);
        for i=1:size(XP,1)
            a=XP(i,1);
            b=YP(i,1);
            if a>0 & a<145 & b>0 & b<145
                mat(a,b)= T(tnum).LoAn(i,2);
            end
        end
     end

[Walter Roberson]

Isn't your RHO the same for each iteration? If so then compute it outside the loop. Same for THETA.

To confirm: LoAn is a 2D array, and not just a row vector ? so size(LoAn,1) > 1 ?

[Roger Stafford]

It looks as though your purpose in this code is to set up the array 'mat'. However, on each trip through the outer for-loop this array is completely erased and refilled with the information from the next 'rotation' angle. What is the sense in that? Only the last 360 degree angle would have any effect on the final values in 'mat'. Is it possible you are doing something with 'mat' which you haven't shown here before returning for another rotation angle? We need to know such things if we are to help you effectively.

Answer 1

[Azzi Abdelmalek]

You can start by pre-allocating

http://www.mathworks.com/help/matlab/matlab_prog/techniques-for-improving-performance.html

[Ronaldo]

In other part of the code, I did pre-allocating. I like to make more parallel calculations. Since I am going to use spmd, I try to avoid parfor loops.