User Guide

>
For more information, see the ?evalf and ?Digits help pages.
Note: When appropriate, Maple performs floating-point computations directly
using your computer's underlying hardware.
Sources of Error
By its nature, floating-point computation normally involves some error.
Controlling the effect of this error is the subject of active research in Numer-
ical Analysis. Some sources of error are:
• An exact quantity may not be exactly representable in decimal form:
and are examples.
• Small errors can accumulate after many arithmetic operations.
• Subtraction of nearly equal quantities can result in essentially no useful
information. For example, consider the computation for
>
No correct digits remain. If, however, you use Maple to analyze this expres-
sion, and replace this form with a representation that is more accurate for
small values of , a fully accurate 10-digit result can be obtained.
>
70 • 3 Performing Computations