如何减轻转置数组访问顺序对性能的影响？

I have an algorithm, written in C, which processes a couple of 2-dimensional arrays (e.g. with size Y x X) to produce another 2-dimensional array of the same size. All three arrays contain 32-bit floats and have the same size, Y x X, where Y might be a few tens but X is a million or so.I have an algorithm, written in C, which proces