For loop optimization in matrix operations

I'm trying to figure out how to optimize 3 nested for loops with matrix operations. The problem boils down to:

for ii = 1:1000
    for jj = 1000
        for kk = 1:100
            P{ii,jj,kk} = [ exp(-1i*klz(ii,jj,kk)*d(kk)), 0 ; ...
                            0, exp(1i*klz(ii,jj,kk)*d(kk)) ];
        end
    end
end

Is there a way to do something like:

for kk = 1:100
    P{:,:,kk} = [ exp(-1i*klz(:,:,kk)*d(kk)), 0 ; ...
                  0, exp(1i*klz(:,:,kk)*d(kk)) ];
end

with some function / method of coding it to decrease the run time?

Answer 1

[Steven Lord]

You don't need to loop to generate the values that you multiply by +1i or -1i and pass into exp.

% Sample data
x = randn(4, 5, 6);
y = 1:6;

% Approach 1: element-wise multiplication
z1 = x.*reshape(y, 1, 1, size(x, 3));

% Approach 2: loops
z2 = zeros(size(x));
for r = 1:size(x, 1)
    for c = 1:size(x, 2)
        for p = 1:size(x, 3)
            z2(r, c, p) = x(r, c, p)*y(p);
        end
    end
end

% Check
z1-z2 % Should contain all 0's or small magnitude values

[Morgan Blankenship]

well exp() will return a 1000x1000 array and then when I try to add it to P it'll throw a cat error because I'm trying to do P = [ [1000x1000], 0; 0, [1000x1000]].

[Morgan Blankenship]

z1 and z2 are supposed to be cells not matrices by the way.