Should we collapse the spatial dimensions to save RAM?

[SamStudio8]

Hansel considers symmetric observations, thus half of the (A^2)*(B^2) matrix (for A symbols and B positions) is strictly zero. We can thus collapse the two spatial dimensions into one dimension:

Z[m][n][i][j] == Z[m][n][i + ((j-1)*(j))/2]

But the question is whether the space overhead is worth the time overhead? (Probably)