A real constant approximates the value of a mathematical real
  number.  The value of the constant can be positive, zero, or
  negative.

  VSI Fortran provides three kind type parameters for data of type
  real:  REAL(KIND=4) (or REAL*4), REAL(KIND=8) (or REAL*8), and
  REAL(KIND=16) (or REAL*16).  REAL(KIND=8) is DOUBLE PRECISION.  If
  DOUBLE PRECISION is used, a kind type parameter must not be
  specified for the constant.

  The type specifier for the real (single-precision) type is REAL;
  the type specifier for the double precision type is DOUBLE
  PRECISION.

  If a kind type parameter is specified, the real constant has the
  kind specified.  If a kind type parameter is not specified, the
  kind is default real.

  The following is the general form of a real constant with no
  exponent part:

    [s]n[n...][_k]

  A real constant with an exponent part has one of the following
  forms:

    [s]n[n...]E[s]nn...[_k]
    [s]n[n...]D[s]nn...
    [s]n[n...]Q[s]nn...

     s   Is a sign; required if negative (-), optional if
         positive (+).

     n   Is a decimal digit (0 through 9). A decimal point
         must appear if the real constant has no exponent part.

     k   Is an optional kind type parameter (4 for REAL(KIND=4),
         8 for REAL(KIND=8), or 16 for REAL(KIND=16)).  It must
         be preceded by an underscore (_).

  Leading zeros (zeros to the left of the first nonzero digit) are
  ignored in counting significant digits.  For example, in the
  constant 0.00001234567, all of the nonzero digits, and none of the
  zeros, are significant.  (See the following sections for the number
  of significant digits each kind type parameter typically has).

  The exponent represents a power of 10 by which the preceding real
  or integer constant is to be multiplied (for example, 1.0E6
  represents the value 1.0 * 10**6).

  A real constant with no exponent part is (by default) a
  single-precision (REAL(KIND=4)) constant.  You can change this
  default behavior by specifying the compiler option
  /ASSUME=FPCONSTANT.

  If the real constant has no exponent part, a decimal point must
  appear in the string (anywhere before the optional kind type
  parameter).  If there is an exponent part, a decimal point is
  optional in the string preceding the exponent part; the exponent
  part must not contain a decimal point.

  The exponent letter E denotes a single-precision real (REAL(KIND=4)
  or REAL*4) constant, unless the optional kind type parameter
  specifies otherwise.  For example, -9.E2_8 is a double-precision
  constant (which can also be written as -9.D2).

  The exponent letter D denotes a double-precision real (REAL(KIND=8)
  or REAL*8) constant.

  The exponent letter Q denotes a quad-precision real (REAL(KIND=16)
  or REAL*16) constant.  A minus sign must appear before a negative
  real constant; a plus sign is optional before a positive constant.
  Similarly, a minus sign must appear between the exponent letter (E,
  D, or Q) and a negative exponent, whereas a plus sign is optional
  between the exponent letter and a positive exponent.

  If the real constant includes an exponent letter, the exponent
  field cannot be omitted, but it can be zero.

  To specify a real constant using both an exponent letter and a kind
  type parameter, the exponent letter must be E, and the kind type
  parameter must follow the exponent part.