Practice Problems: Chapter 1
1. The floating point representation can be expressed in any of the following forms:
Standard Normalized Form : 𝐹 = (± )β × β𝑒;
0. 𝑑1𝑑2𝑑3 ··· 𝑑𝑚 𝑑1 ≠ 0.
IEEE Normalized Form : 𝐹 = (± 0. 1𝑑1𝑑2𝑑3 ··· 𝑑𝑚 )β × β𝑒 .
IEEE Denormalized Form : 𝐹 = (± )β × β𝑒 .
1. 𝑑1𝑑2𝑑3 ··· 𝑑𝑚
a) Consider a system with β = 2, 𝑚 = 4, and − 3 ≤ 𝑒 ≤ 4. Find out the maximum and
minimum numbers this system can store with and without negative support. Express the
numbers both in binary and decimal digits for all three forms.
b) How many numbers can this system represent or store in all these forms?
c) Using Standard Form, find all the decimal numbers without negative support, plot them
on a real line, and show if the number line is equally spaced or not.
d) For the IEEE standard for double-precision (64-bit) arithmetic, find the smallest positive
number and the largest number representable by a system that follows this standard. Do
not find their decimal values, but simply represent the numbers in the following format:
(± 0. 1𝑑1𝑑2𝑑3 ··· 𝑑𝑚)β × β𝑒−𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝐵𝑖𝑎𝑠 .
Be mindful of the conditions for representing ±∞ and ±0 in this IEEE standard.
e) In the above IEEE standard, if the exponent bias were to be altered to
𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡 𝐵𝑖𝑎𝑠 = 500, what would the smallest positive number and the largest number
be? Write your answers in the same format as in part (d). Note that the conditions for
representing ±∞ and ±0 are still maintained as before.
2. If 𝑥 = 3/8 and 𝑦 = 5/8, find 𝑓𝑙(𝑥 × 𝑦) where 𝑚 = 4. Also check whether
𝑥 × 𝑦 = 𝑓𝑙(𝑥 × 𝑦). If not, find the rounding error of the product of these two numbers.
2
3. Consider the quadratic equation, 𝑥 − 60𝑥 + 1 = 0 . Working to 6 significant figures,
compute the roots of the quadratic equation and check that there is a loss of significance.
Find the correct roots such that loss of significance does not occur.
4. Given β = 2, 𝑚 = 5 , − 100 ≤ 𝑒 ≤ 100. Using the IEEE Normalized form, answer the
following:
a) Compute the Machine Epsilon (𝞊M).
b) Compute the minimum of ∣x∣.
c) How many non-negative numbers can you represent using this system?
2
5. Consider the quadratic equation 𝑥 − 16𝑥 + 3 = 0. Explain how the loss of significance
occurs in finding the roots of the quadratic equation if we restrict to 4 significant figures.
Discuss how to avoid this and find the roots.
6. Given a system parameterized by 𝛽 = 2, 𝑚 = 3, and 𝑒min = −1 ≤ 𝑒 ≤ 𝑒max = 2, where
𝑒 ∈ 𝑍. For this system answer the following:
(a) Find the floating-point representation of the numbers (6.25)10 and (6.875)10 in the
Normalized Form. That is, find 𝑓𝑙(6. 25)₁₀ and 𝑓𝑙(6. 875)₁₀ .
(b) What are the rounding errors 𝛿1,𝛿2 in part (a)?
(c) Can the values (6.25)10 and (6.875)10 be represented in the Denormalized Form? If so,
find the floating-point representations. If not, then concisely explain why?
(d) Find the rounding error for Standard Form, Normalized and Denormalized Form.
7. Consider the real number x = (8.235)10
(a) First convert the decimal number x in binary format at least up to 8 binary places.
(b) What will be the binary value of x [Find fl(x)] if you store it in a system with m = 6 using
the Denormalized form of floating point representation.
(c) Now convert back to decimal form the stored values you obtained in the previous part,
and calculate the rounding error of both numbers.
8. Consider the quadratic equation:
2
𝑥 − 12𝑥 + 5 = 0
a) Compute the roots of the quadratic equation while keeping to four significant figures.
b) Explain how loss of significance occurs in this case due to the subtraction of nearly
equal numbers.
c) Discuss an alternative approach to computing the roots to avoid loss of significance,
and use this method to determine the correct roots.
9. Consider a computing system with base 𝛽 =2, 𝑚 = 3, and 𝑒min = −3 ≤ 𝑒 ≤ 𝑒max = 2
a) In the Standard form of this system, determine the total number of representable values
including support for negative numbers. Also, compute the maximum value of delta.
b) Express the floating-point representations (binary format) for the numbers 𝑥 = 4/8 and
𝑦 = 7/8 in this system.
c) Compute 𝑓𝑙(𝑥 × 𝑦) and determine whether this value can be stored within the given
floating-point system.
