Generally speaking, you may use a floating-point data item anywhere that you can use a non-integer data item. Data moved to or from a floating-point item is converted to the appropriate format.
If you use a floating-point item in an arithmetic expression, then that expression is computed by converting all the values to double-precision floating-point and doing the arithmetic using the machine's conventions for double-precision math. The result is then converted to the type appropriate for the destination.