Unformatted text preview: hat is that Precision?
= 2;23 1:2 10;7
= 2;52 1:1 10;16
IEEE Extended = 2;63 1:1 10;19
exact machine precision, together with its approximate decimal equivalent, is
shown in Table 3 for each of the IEEE single, double and extended formats.
The fraction of a single precision normalized number has exactly 23 bits of
accuracy, i.e. the signi cand has 24 bits of accuracy counting the hidden bit.
This corresponds to approximately 7 decimal digits of accuracy. In double
precision, the fraction has exactly 52 bits of accuracy, i.e. the signi cand has
53 bits of accuracy counting the hidden bit. This corresponds to approximately 16 decimal digits of accuracy. In extended precision, the signi cand
has exactly 64 bits of accuracy, and this corresponds to approximately 19 decimal digits of accuracy. So, for example, the single precision representation
for the number is approximately 3:141592, while the double precision representation is approximately 3:141592653589793. To see exactly what the...
View Full Document
- Spring '09
- oating point