A floating-point item is a numeric data item that allows for a very wide range of values. However, compared to other numeric data types in COBOL, floating-point data is less accurate. Most computer languages use floating-point to represent non-integer values. This makes floating-point a good method for sharing non-integer data with these other languages.
Floating-point data items differ in several ways from normal numeric data items:
Floating-point items are stored in a machine-dependent format. In particular,
they are stored in a format that is "native" to each machine. There are many floating-point formats currently in use by
different machines, so floating-point data should not be considered portable.
Floating-point items do not have a picture associated with them. Instead,
floating-point items are either 4 or 8 bytes in size. The size of the item
determines the range of values it can hold.
The range of values that can be stored in a floating-point item is
machine-dependent.
Because floating-point items do not maintain accuracy very well, you should limit their use. Some examples where floating-point is appropriate are:
You need to share non-integer data with another language such as C or FORTRAN.
You need to hold very large or very small values that exceed the usual 18
digits available in COBOL.
You need to process existing data that contains floating-point values.