Why does fsamp2 perform fftshift on output coefficients?

Why does matlab internal function fsamp2 perform fftshift on the output filter coefficients? After swapped, the output cannot be a direct input to filter2 function, right?

Following are an example and the concerned part of codes in fsamp2.m.

% For example, to create a low-pass filter LPF with frequency response Hd
[f1,f2] = freqspace(25,'meshgrid');
Hd = zeros(25,25); d = sqrt(f1.^2 + f2.^2) < 0.5; % (cutoff at 0.5)
Hd(d) = 1;

% Here's my thought of creating a filter:
LPF = fftshift(Hd); % convolution kernel
LPF = rot90(LPF,2); % correlationn kernel
signal_filtered = filter2(LPF,signal);

% But if I use fsamp2:
h = fsamp2(Hd);
% According to the fsamp2 function below (mainly the line commented with "QUESTION"), 
% h is not the same with LPF and cannot be a direct input to filter2 function.
% However, h can be used with freqz2(h) directly because freqz2 swaps h back.

function h = fsamp2(f1,f2,hd,siz)

if nargin==1, % Uniform spacing case (fast)
  hd = f1;

  hd = rot90(fftshift(rot90(hd,2)),2); % Inverse fftshift
  h = fftshift(ifft2(hd)); % QUESTION: ifft2(hd) already outputs the desired coefficients, why use fftshift then?
elseif nargin==2 || nargin==3,
    % etc...
else
    % etc...
end
% etc
h = rot90(h,2); % Rotate for use with filter2

end % end function

[Sulaymon Eshkabilov]

It is used to swap the right halves of your computed vector values.

[skykeithtr]

Yes, but I just can't see why it needs to be swapped. I have edited the question to discuss whether it would be an appropriate input to other functions like filter2 and imfilter.