Representing and Manipulating

Information
Floating-Point Number Representation
 A floating-point number (or real number) can represent a very large value
(e.g., 1.23×10^88)
 or a very small value (e.g., 1.23×10^-88).
 It could also represent very large negative number (e.g., -1.23×10^88)
 and very small negative number (e.g., -1.23×10^-88), as well as zero, as
illustrated:
Floating-Point Number Representation
 A floating-point number is typically expressed in the scientific notation,
 With a fraction (F), and an exponent (E) of a certain radix (r), in the form of
F×r^E.
 Decimal numbers use radix of 10 (F×10^E);
 While binary numbers use radix of 2 (F×2^E).
 Representation of floating point number is not unique.
 For example, the number 55.66 can be represented as 5.566×101,
0.5566×102, 0.05566×103, and so on.
Floating-Point Number Representation
 The fractional part can be normalized.
 In the normalized form, there is only a single non-zero digit before the
radix point.
 For example, decimal number 123.4567 can be normalized as
1.234567×102;
 binary number 1010.1011B can be normalized as 1.0101011B×23.
Floating-Point Number Representation
 It is important to note that floating-point numbers suffer from loss of
precision.
 When represented with a fixed number of bits (e.g., 32-bit or 64-bit).
 This is because there are infinite number of real numbers (even within a
small range of says 0.0 to 0.1).
 On the other hand, a n-bit binary pattern can represent a finite 2n distinct
numbers.
 Hence, not all the real numbers can be represented.
 The nearest approximation will be used instead, resulted in loss of
accuracy.
Floating-Point Number Representation
 Floating number arithmetic is very much less efficient than integer
arithmetic.
 It could be speed up with a so-called dedicated floating-point co-
processor.
 Hence, use integers if your application does not require floating-point
numbers.
Floating-Point Number Representation
 In computers, floating-point numbers are represented in scientific
notation of fraction (F) and exponent (E) with a radix of 2, in the form of
F×2^E.
 Both E and F can be positive as well as negative.
 Modern computers adopt IEEE 754 standard for representing floating-
point numbers.
 There are two representation schemes: 32-bit single-precision and 64-bit
double-precision.
IEEE-754 32-bit Single-Precision Floating-Point
Numbers
 In 32-bit single-precision floating-point representation:
 The most significant bit is the sign bit (S),
 with 0 for positive numbers and 1 for negative numbers.
 The following 8 bits represent exponent (E).
 The remaining 23 bits represents fraction (F).
Normalized Form
 Let's illustrate with an example, suppose that the 32-bit pattern is,
 1 1000 0001 011 0000 0000 0000 0000 0000, with:
 S=1
 E = 1000 0001
 F = 011 0000 0000 0000 0000 0000
Normalized Form
 In the normalized form, the actual fraction is normalized with an implicit
leading 1 in the form of 1.F.
 In this example, the actual fraction is 1.011 0000 0000 0000 0000 0000 = 1
+ 1×2-2 + 1×2-3 = 1.375D.
 The sign bit represents the sign of the number,
 with S=0 for positive and S=1 for negative number.
 In this example with S=1, this is a negative number, i.e., -1.375D.
Normalized Form
 The exponent field is interpreted as representing a signed integer in biased
form.
 That is, the exponent value is E = e − Bias,
 where e is the unsigned number having bit representation ek−1 . . . e1e0
 and Bias is a bias value equal to 2k−1 − 1.
 This yields exponent ranges from −126 to +127.
Normalized Form
 Why set the bias this way for denormalized values?
 Having the exponent value be 1 − Bias rather than simply −Bias.
 it provides for smooth transition from denormalized to normalized values.
Normalized Form
 In this example, E=e-127=129-127=2D.
 Hence, the number represented is -1.375×22=-5.5D.
De-Normalized Form
 Normalized form has a serious problem, with an implicit leading 1 for the
fraction,
 it cannot represent the number zero.
 When the exponent field is all zeros, the represented number is in
denormalized form.
 In this case, the exponent value is E = 1 − Bias.
 The value of the fraction field without an implied leading 1.
De-Normalized Form
 Denormalized numbers serve two purposes.
 First, they provide a way to represent numeric value 0,
 Since with a normalized number we must always have F ≥ 1, and hence we
cannot represent 0.
 In fact, the floating-point representation of +0.0 has a bit pattern of all
zeros: the sign bit is 0,
 the exponent field is all zeros (indicating a denormalized value), and the
fraction field is all zeros, giving F = 0.
De-Normalized Form
 when the sign bit is 1, but the other fields are all zeros, we get the value
−0.0.
 A second function of denormalized numbers is to represent numbers that
are very close to 0.0.
De-Normalized Form
 We can also represent very small positive and negative numbers in de-
normalized form with E=0.
 For example, if S=1, E=0, and F=011 0000 0000 0000 0000 0000.
 The actual fraction is 0.011=1×2-2+1×2-3=0.375D.
 Since S=1, it is a negative number.
 With E=0, the actual exponent is -126.
 Hence the number is -0.375×2-126 = -4.4×10-39,
 which is an extremely small negative number (close to zero).
Special Values
 A final category of values occurs when the exponent field is all ones.
 When the fraction field is all zeros, the resulting values represent infinity,
 either +∞ when s = 0 or −∞ when s = 1.
 Infinity can represent results that overflow, as when we multiply two very
large numbers, or when we divide by zero.
 When the fraction field is nonzero, the resulting value is called a NaN,
short for “not a number.”
IEEE-754 64-bit Double-Precision Floating-
Point Numbers
 The representation scheme for 64-bit double-precision is similar to the 32-
bit single-precision:
 The most significant bit is the sign bit (S), with 0 for positive numbers and
1 for negative numbers.
 The following 11 bits represent exponent (E).
 The remaining 52 bits represents fraction (F).
IEEE-754 64-bit Double-Precision Floating-
Point Numbers
 The value (N) is calculated as follows:
 Normalized form: For 1 ≤ E ≤ 2046, N = (-1)^S × 1.F × 2^(E-1023).
 Denormalized form: For E = 0, N = (-1)^S × 0.F × 2^(-1022). These are in
the denormalized form.
 For E = 2047, N represents special values, such as ±INF (infinity), NaN (not
a number).