The FLOAT attribute indicates the type of precision to  use  for
  objects of type REAL and DOUBLE.

  Syntax:

     [FLOAT(keyword)]

  You can specify the following float types for keywords:

   o  D_FLOAT - REAL=FLOATING, DOUBLE=D_FLOATING

   o  G_FLOAT - REAL=F_FLOATING, DOUBLE=G_FLOATING

   o  IEEE_FLOAT - REAL=S_FLOATING, DOUBLE=T_FLOATING

  The [FLOAT] attribute is only valid on compilation-units.