Reducing the computation time of the planar antenna array pattern calcuation

Hi, i'm calculating the planar antenna array pattern, the problem is that it takes ~10s to calculate the pattern, and i need to calculate it a lot of times in my algorithm. The majority of time is spent to calculate the for loop. Please advise me how can i reduce the for loop execution time, or replace it with some other solution. Thanks in advance.

    clc; close all; clear all;

Nx=32; Ny=32;
betta_n = zeros(Nx,Ny); lambda=1;
dx = [1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 ];
dy = [1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 ];
a_n = ones(Nx, Ny);
dtetta = 0.01;
dphi = 0.01;

tetta0 = -pi/2:dtetta:pi/2;
phi0 = dphi:dphi:2*pi;
[phi, tetta] = meshgrid(phi0,tetta0);
j = sqrt(-1);
amp_phase=a_n.*(cos(betta_n)+j*sin(betta_n));

sinU=sin(tetta).*cos(phi);
sinV=sin(tetta).*sin(phi);
%     AF_Field = 0;
AF_Field = zeros(length(tetta0),length(phi0));
tic
for n=1:Nx-1
    for m=1:Ny-1
        %calculate each of the exponents separately wo the loop
        AF_Field = amp_phase(n,m)*exp(j*2*pi*(n-1)/lambda*dx(n)*(sinU)).*exp(j*2*pi*(m-1)/lambda*dy(m)*(sinV)) + AF_Field;  
    end
end
toc

Answer 1

[Chandra Amma Palanisamy]

The main computational cost comes from repeatedly evaluating the exponential terms over the entire (θ, ϕ) grid inside the nested loops. To improve performance, precompute the x- and y-direction phase terms and reuse them, or express the array factor as a matrix operation to eliminate the double loop.

Additional options:

• Cache the phase terms if the array geometry is fixed.
• Use parfor or GPU acceleration for large optimization runs.
• Use the MATLAB Profiler to identify remaining bottlenecks.