Section 8,4,8: DOUBLE_PRECISION

 A DOUBLE PRECISION (REAL*8) constant can be a basic real constant
 or an integer constant followed by a decimal exponent.  A decimal
 exponent has the form:

    Dsnnn        s is an optional sign
                 nnn is a string of decimal digits

 NOTE:  DEC Fortran also allows the syntax Qsnnn, if the exponent
 field is within the T_floating double precision range.

 On RISC processors, the DOUBLE PRECISION data type implements the
 IEEE T_floating format.  The approximate ranges of a nonzero DOUBLE
 PRECISION constant are 2.225D-308 to 1.798D308 (T_floating).

 T_floating computational results smaller than
 2.2250738585072012D-308 are denormalized numbers that gradually
 underflow and lose precision, until they reach approximately
 4.9406564584124654D-324.  You cannot write a constant for a
 denormalized number.