• A should be a square matrix, i.e., ndims(A) == 2 and size(A,1) == size(A,2).
• If A is non-singular, inv(A) gives its matrix inverse.
• It returns an error if A is singular.
• It reports a warning, and proceeds to compute the inverse anyway, if the reciprocal of the condition number of A is less than norm(A)*eps (i.e., A nearly singular or badly conditioned).