Modified Floating-Point Arithmetic
Original Publication Date: 1985-Oct-01
Included in the Prior Art Database: 2005-Feb-19
In a modified floating-point arithmetic context, data which falls in the standard range should be expressed essentially in fixed point, using floating-point format but with a standard exponent. Other data would be expressed in the normal floating-point way. The advantages of floating-point arithmetic are well known, [*]. The disadvantages, in terms of performance, in those cases where fixed point would be adequate, are also well known. They derive largely from the need to shift arguments for alignment before addition and subtraction, and to normalize results. In many contexts, e.g., Plancode-like programs or spreadsheets, much of the arithmetic involves numbers that could be expressed with at most twelve digits before the decimal point and five after. Such data could conventionally be handled in some fixed-point format.