Unformatted text preview: in this case, b23 b24 : : : are all zero. For example, the number
11=2 = (1:011)2 22
would be represented by
0 E=2 1.0110000000000000000000
1 and the number
would be represented by
0 71 = (1:000111)2 26 E=6 1.0001110000000000000000 : To avoid confusion, the exponent E , which is actually stored in a binary
representation, is shown in decimal for the moment.
The oating point representation of a nonzero number is unique as long
as we require that 1 m < 2. If it were not for this requirement, the number
11=2 could also be written
(0:01011)2 24
and could therefore be represented by
0 E=4 0.0101100000000000000000 : However, this is not allowed since b0 = 0 and so m < 1. A more interesting
example is
1=10 = (0:0001100110011 : : :)2:
Since this binary expansion is in nite, we must truncate the expansion somewhere. (An alternative, namely rounding, is discussed later.) The simplest
way to truncate the expansion to 23 bits would give the representation
0 E=0 0.0001100110011001100110 but this means m < 1 since b0 = 0. An eve...
View
Full Document
 Spring '09
 LIE
 oating point

Click to edit the document details