You are on page 1of 7

B

FLOATING-POINT NUMBERS

3 Negative underflow 1 Negative overflow 2 Expressible negative numbers 4 Zero

5 Positive underflow 6 Expressible positive numbers 7 Positive overflow

10100

10100 0

10100

10100

Figure B-1. The real number line can be divided into seven regions.

Digits in fraction Digits in exponent Lower bound Upper bound 3 3 3 3 4 4 4 4 5 5 5 5 10 20 1 2 3 4 1 2 3 4 1 2 3 4 3 3 1012 10102 101002 1010002 1013 10103 101003 1010003 1014 10104 101004 1010004 101009 101019 109 1099 10999 109999 109 1099 10999 109999 109 1099 10999 109999 10999 10999

Figure B-2. The approximate lower and upper bounds of expressible (unnormalized) floating-point decimal numbers.

Example 1: Exponentiation to the base 2


22 2
1

24 2
3

26 2
5

28 2
7

210 2
9

212
11

214
13

216
15

Unnormalized:

0 1010100

+ 1 216) = 432 Sign Excess 64 Fraction is 1 212+ 1 213 15 16 + exponent is +1 2 + 1 2 84 64 = 20 To normalize, shift the fraction left 11 bits and subtract 11 from the exponent. Normalized: 0 1001001 Sign Excess 64 + exponent is 73 64 = 9

.0

20 12 13 15 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 = 2 (1 2 + 1 2 + 1 2

.1

1 0 1 1 0 0 0 0 0 0 0 0 0 0 0 = 29 (1 21+ 1 22+ 1 24 Fraction is 1 21 + 1 22 +1 24 + 1 25 Example 2: Exponentiation to the base 16 161 162 0 0 00 163 0 0 01 164 1 0 1 1 = 165 (1 163+ B 164) = 432 + 1 25) = 432

Unnormalized:

0 1000101 Sign Excess 64 + exponent is 69 64 = 5

0 0 00

Fraction is 1 163 + B 164

To normalize, shift the fraction left 2 hexadecimal digits, and subtract 2 from the exponent. Normalized: 0 1000011 Sign Excess 64 + exponent is 67 64 = 3

0001

1011

0000

0 0 0 0 = 163 (1 161+ B 162) = 432

Fraction is 1 161 + B 162

Figure B-3. Examples of normalized floating-point numbers.

Bits 1

23 Fraction

Sign

Exponent (a)

Bits 1

11 Exponent Sign (b)

52 Fraction

Figure B-4. IEEE floating-point formats. (a) Single precision. (b) Double precision.

Item Bits in sign Bits in exponent Bits in fraction Bits, total Exponent system Exponent range Smallest normalized number Largest normalized number Decimal range Smallest denormalized number

Single precision 1 8 23 32 Excess 127 126 to +127 2126 approx. 2128 approx. 1045

Double precision 1 11 52 64 Excess 1023 1022 to +1023 21022 approx. 21024 approx. 10324

approx. 1038 to 1038 approx. 10308 to 10308

Figure B-5. Characteristics of IEEE floating-point numbers.

Normalized Denormalized Zero Infinity Not a number

0 < Exp < Max 0 0 1 1 11 1 1 11 Sign bit

Any bit pattern Any nonzero bit pattern 0 0 Any nonzero bit pattern

Figure B-6. IEEE numerical types.

You might also like