Numerical implementation of diagonalization solutions

Numerical procedures for the diagonalization and eigenvector calculation of matrices all trace back to EISPACK, a freely available software collection of subroutines originally written in double-precision FORTRAN77, but now also available as 

3/2 energy matrix of Equation 7.38 into the form given in Equation 8.29, and for the latter we worked out the terms pp St q for the intensity expressions. This switching of matrix form holds a warning the coefficients of the worked-out (pp S, q terms 

For matrices of modest dimensions 1024 matrix diagonalizations may not be a serious CPU problem for a PC, but if we include (as we will in the next chapter) distributions in the spin Hamiltonian parameters the required CPU time goes up by, say, two orders of magnitude, and if we want to implement automatic minimization, we must pay with another two or three orders of magnitude in CPU-time. 

---