! for portable 64-bit floating-point precision [via REAL(q), COMPLEX(q), ...]
      MODULE PREC
      INTEGER, PARAMETER :: q=SELECTED_REAL_KIND(10)
      END MODULE