Roots of a Single-Variable  Representing Numbers  Overflow and Underflow

Loss of Significance Errors

The fixed-length representation of floating-point numbers can also bring about unexpected results in very simple arithmetic operations without causing underflow or overflow errors. While one cannot possibly circumvent all possible scenarios where this might occur, there are certain programming techniques which can minimize their influence. The most notable situation where the fixed-length representation is problematic appears when computing the difference of nearly-identical numbers. This error is also referred to as subtractive cancellation.

 

Wed Oct 28 11:42:13 EST 1998