How to parallelize this code or make it more efficient

I'm doing 3D fft and I would like to parallelize this code if possible, or at least make it more efficient.

Dim=size(y);
N1_FFT=2*2^nextpow2(Dim(1));
N2_FFT=2*2^nextpow2(Dim(2));
N3_FFT=2*2^nextpow2(Dim(3));

N_FFT=[N1_FFT,N2_FFT,N3_FFT];

%Calculate 3D fft
CostFunction3=y;
for p = 1:length(size(y))
    CostFunction3 = fft(CostFunction3,N_FFT(p),p);
end
CostFunction3 = fftshift(abs(CostFunction3).^2);

Edit: forgot to mention, y is a 3D matrix with signal samples

[Adam]

If y is a vector how is this a 3D FFT?

Also you should use

ndims(y)

if you want to loop over the number of dimensions, which is what length( size(y) ) will give you, but in your case, if y is a vector this will return 2 which is not helpful really anyway because one of those dimensions will be a singleton.

[Omar Hammouda]

Sorry for not being clear, y is a 3D matrix 256x256x256

also I forgot to add this part for calculating the N_FFT

Dim=size(y);
N1_FFT=2*2^nextpow2(Dim(1));
N2_FFT=2*2^nextpow2(Dim(2));
N3_FFT=2*2^nextpow2(Dim(3));

N_FFT=[N1_FFT,N2_FFT,N3_FFT];

I've fixed my question, thanks for noting the mistakes