Unformatted text preview: E 2E =1:
2 x is subnormal, i.e. E = ;126 and b0 = 0? Now another way to write the de nition of is
round(x) = x(1 + )
so we have the following result: the rounded value of a normalized number x
is, when not exactly equal to x, equal to x(1 + ), where, regardless of the
rounding mode, j j< : Here, as before, is the machine precision. In the case of round to nearest,
we have
1
jj 2:
This result is very important, because it shows that, no matter how x is
displayed, for example either in binary format or in a converted decimal
format, you can think of the value shown as not exact, but as exact within a
factor of 1 + . Using Table 3 we see, for example, that IEEE single precision
numbers are good to a factor of about 1 + 10;7, which means that they have
about 7 accurate decimal digits.
Numbers are normally input to the computer using some kind of highlevel programming language, to be processed by a compiler or an interpreter.
There are two di erent ways that a number such as 1=10 might be input. One
way wo...
View
Full Document
 Spring '09
 LIE
 oating point

Click to edit the document details