Why is Gauss elimination taking so much time in this algorithm?

Following is a part of an algorithm i wrote for formulating block tri diagonal matrix to solve an elliptic pde via 5 point stencil formula and natural ordering:

M = 4;
N = M;
h = 1/M;
M1 = (M-1)*(N-1);

F = zeros(M1,1);

h1 = h^2;

x = zeros(N-1,1);

for i = 1:N-1
    x(i)  = h*i;
end

y = zeros(M-1,1);
for j = 1:M-1
    y(j)  = h*j;
end

for j = 1:N-1
    for i = 1:M-1
        k = i + (M-1)*(j-1);
        F(k) = f(x(i),y(j));
    end
end

% define block B%

B = zeros(N-1,N-1);

B(1,1) = -4/h1;
B(1,2) = 1/h1;
B(2,1) = 1/h1;

for k = 2:N-2

      B(k,k) = -4/h1;
      B(k+1,k) = 1/h1;
      B(k,k+1) = 1/h1;
  end

B(N-1,N-2) = 1/h1;
B(N-1,N-1) = -4/h1;

% Concatenate A
blocks = repmat({B}, N-1, 1);
res = blkdiag(blocks{:});

R = repmat(1/h1,1,M1-(M-1));
I1 = diag(R,M-1);
I2 = diag(R,-(M-1));

A = res+I1+I2;

tic
 U = A\F; 
toc

f(x,y) is defined in a function file as follows,

function W = f(x,y)

W = -2*pi^2*sin(pi*x)*sin(pi*y);

end

For large values of M gauss elimination takes a lot of time. For instance when i take M = 100 the G elimination step alone takes precisely 25.499652 seconds.

I cant figure out why gauss elimination is taking so much time. Am i missing some initializations? Any suggestions will be helpful.

Regards