This preview shows page 1. Sign up to view the full content.
Unformatted text preview: to zero or subnormal number
Set result to correctly rounded value
normalized numbers? 5 .
Exercise 24 Consider the computation of the previous exercise again. Is it possible that d and e could have many less digits of accuracy than a and b,
even though a and b are normalized numbers? Altogether, the IEEE standard de nes ve kinds of exceptions: invalid
operation, division by zero, over ow, under ow and precision, together with
a standard response for each of these. All of these have just been described
except the last. The last exception is, in fact, not exceptional at all because it
occurs every time the result of an arithmetic operation is not a oating point
number and therefore requires rounding. Table 4 summarizes the standard
response for the ve exceptions.
The IEEE standard speci es that when an exception occurs it must be
signaled by setting an associated status ag, and that the programmer should
have the option of either trapping the exception, providing special code to
be executed when the exception occurs, or masking the exception, in which
case the program continues executing with the standard response shown in
the table. If the user is not programming...
View Full Document
- Spring '09