Fortran Real Double Precision