A discussion of the consequences that can arise from naive usage of symbolic computation when generating Fortran code. Given the following 3 by 3 symmetric matrix M (perhaps created in a symbolic computation system like Maple), and we want to evaluate numerically the inverse of M at particular values of the parameters q2, q3, p, m10, m30, j10y, j30x, j30y, j30z in a Fortran program. It is tempting to compute the inverse of the matrix M symbolically then use Maple's fortran function to generate the Fortran code. Does this produce the most efficient code? Let's see.