Optimizing a Sliding Window for Matrix Re-Assignment

I have a matrix, A, which is a massive NxM matrix where N is several millions and M is relatively small. and I would like to do an operation (xcorr() or std(), for example) on a window of this matrix, i:i+x, for all i. However, using for loops ends up taking a lot of time since Matlab and for loops get along like me and my ex wife.

I'm trying to optimize it by creating a new matrix, B composed of the windows of A. For example, if

A=[1 2 3 4 5 6 8 9 10],

and the window size is x=3,

B= [1 2 3
    2 3 4
    3 4 5
    ...
    8 9 10]

Is there any way to optimize this using only vector/matrix operations in order to speed up processing?