Integer representations and Algorithms – Rosen

Section 4.2

Anna-Simone Frank1

MNF130

Spring 2024

1
Slides created by Tom Michoel and modified by Erik Maartensson and
Anna-Simone Frank
Integer representations

Definition
If b > 1 and n > 0 are integers, then the base b expansion of n is the
unique expression of n in the form

n = ak b k + ak−1 b k−1 + · · · + a1 b + a0

where k is a nonnegative integer, a0 , a1 , . . . , ak are nonnegative integers

less than b, and ak 6= 0. We write n = (ak , ak−1 , . . . , a1 , a0 )b

Example
I The usual decimal expansion is obtained by setting the base
b = 10. Here the choice of basis is normally implied.
I The binary expansion of an integer is obtained by setting the base
b = 2. Then each digit ai is either 0 or 1.
I The octal and hexadecimal expansion are obtained by setting the
base b = 8 and 16, respectively.

4
Example
The decimal representation of the integer with binary expansion
(101100011)2 is:

(10110011)2
= 1 · 27 + 0 · 26 + 1 · 25 + 1 · 24 + 0 · 23 + 0 · 22 + 1 · 21 + 1 · 20
= 128 + 32 + 16 + 2 + 1
= 179

5
Example
The binary representation of the integer 67 (decimal representation) is:

67 mod 2 = 1 67 div 2 = 33
33 mod 2 = 1 33 div 2 = 16
16 mod 2 = 0 16 div 2 = 8
8 mod 2 = 0 8 div 2 = 4
4 mod 2 = 0 4 div 2 = 2
2 mod 2 = 0 2 div 2 = 1
1 mod 2 = 1 1 div 2 = 0

Hence 67 = (1000011)2

6
Decimal from and to octal

(1337)8 = 1 · 83 + 3 · 82 + 3 · 8 + 7 = 597
Conversely: given 1337 in decimal representation.

1337 mod 8 = 1 1337 div 8 = 167

167 mod 8 = 7 167 div 8 = 20
20 mod 8 = 4 20 div 8 = 2
2 mod 8 = 2 2 div 8 = 0

Thus 1337 = (2471)8

7
Hexadecimal representation

In hexadecimal representation we let b = 16 and

k
X
n= ai 16i , (1)
i=0

where ai ∈ {0, 1, . . . , 15}.

We normally only have 10 symbols to denote digits with, so in
hexadecimal we let A = 10, B = 11, C = 12, D = 13, E = 14, F = 15.
We write 10 · 162 + 5 · 16 + 15 = (A5F )16 , for example.

8
Decimal from and to hexadecimal

(1337)16 = 1 · 163 + 3 · 162 + 3 · 16 + 7 = 4919.

Conversely: Given 421337 in decimal representation

421337 mod 16 = 9 421337 div 16 = 26333

26333 mod 16 = 13 26333 div 16 = 1645
1645 mod 16 = 13 1645 div 16 = 102
102 mod 16 = 6 102 div 16 = 6
6 mod 16 = 6 6 div 16 = 0

Thus 421337 = (66DD9)16 .

9
Binary to and from octal

Conversion between binary and octal representations is easy!

I Binary to octal:
I Given (11, 110, 101, 100)2 .
I Zero-pad the number so the length becomes a multiple of 3:
((0)11, 110, 101, 100)2 .
I Translate 3 bits at a time: ((0)11, 110, 101, 100)2 = (3654)8 .
I Octal to binary:
I Given (3654)8 .
I Translate each digit to 3 bits: (3654)8 = ((0)11, 110, 101, 100)2
I Remove the starting zero(s):
((0)11, 110, 101, 100)2 → (11, 110, 101, 100)2

10
Binary to and from hexadecimal

Conversion between binary and hexadecimal representations is easy!

I Binary to hexadecimal:
I Given (111, 1010, 1100)2 .
I Zero-pad the number so the length becomes a multiple of 4:
((0)111, 1010, 1100)2 .
I Translate 4 bits at a time: ((0)111, 1010, 1100)2 = (7AC )16 .
I Hexadecimal to binary:
I Given (7AC )16 .
I Translate each digit to 4 bits: (7AC )16 = ((0)111, 1010, 1100)2
I Remove the starting zero(s):
((0)11, 110, 101, 100)2 → (11, 110, 101, 100)2

11
The general algorithm for constructing Base b expansions is:
procedure base b expansion(b > 1, n : integers)
q := n
k := 0
while q 6= 0 do
ak := q mod b
q := q div b
k := k + 1
end while
return (ak−1 , . . . , a1 , a0 )
end procedure

12
Addition of integers in binary notation
I Follow the usual procedure for decimal representations!

Figure 1: Addition of binary integers

I In pseudocode:
procedure Add(a = (an−1 . . . a1 a0 )2 , b = (bn−1 . . . b1 b0 )2 )
c := 0
for j := 0, . . . , n − 1 do
sj := (aj + bj + c) mod 2
c := (aj + bj + c) div 2
end for
sn := c
return (sn . . . s1 s0 )2 binary expansion of a + b
end procedure

13
Multiplication of integers in binary notation
I Follow the usual procedure for decimal representations!

Figure 2: Multiplication of binary integers

I Let a = (an−1 . . . a1 a0 )2 . First note that for all j

a2j = (a0 20 + a1 21 + · · · + an−1 2n−1 )2j

= a0 2j + a1 2j+1 + · · · + an−1 2j+n−1
= (an−1 . . . a1 a0 0| .{z
. . 0})2
j×

Hence multiplying a by a power of 2 is the same as padding extra

zeros at the tail of the binary expansion.

14
I For a = (an−1 . . . a1 a0 )2 , b = (bn−1 . . . b1 b0 )2 , this can be used as
follows:

ab = a(b0 20 + b1 21 + · · · + bn−1 2n−1 )

= (ab0 )20 + (ab1 )21 + · · · + (abn−1 )2n−1

I If bj = 0, abj = 0. If bj = 1, abj = a. Hence,

X X
ab = a2j = (an−1 . . . a1 a0 0| .{z
. . 0})2 ,
{j : bj 6=0} {j : bj 6=0} j×
P
where the second means using the addition algorithm for binary
expansions.

15
I In pseudocode:
procedure Multiply(a = (an−1 . . . a1 a0 )2 , b = (bn−1 . . . b1 b0 )2 )
for j := 0, . . . , n − 1 do
if bj = 1 then
cj := (an−1 . . . a1 a0 |0 .{z
. . 0})2
j×
else
cj := 0
end if
end for
p := 0
for j := 0, . . . , n − 1 do
p := add(p, cj )
end for
return p binary expansion of ab
end procedure

16
Complexity of binary addition and multiplication

I Adding two n-bit numbers takes O(n) bit operations.

I Multiplying two n-bit numbers means adding n-bit numbers n − 1
times, giving an O(n2 ) algorithm.
I For smaller numbers, on modern computers adding and multiplying
normal-sized numbers take roughly the same time.
I In cryptography we sometimes work with huge numbers, where
the difference between addition and multiplication is really
noticeable.
I Perhaps surpricingly, it’s possible to multiply two n-bit numbers in as
low as O(n log(n)).2

2 https://en.wikipedia.org/wiki/Multiplication_algorithm#

Computational_complexity_of_multiplication 17
Modular exponentiation

I In cryptography we need to be able to compute b n mod m where n

can be very large. Even if n is not large, b n quickly becomes
huge. For instance
367 = 9.2709 × 1031

I Instead of computing b n and then taking the mod operation, we

express n in its binary expansion

n = ak−1 2k−1 + · · · + a1 2 + a0

and write
k−1
b n = b ak−1 2 . . . b a1 2 b a0

where each aj is either 0 or 1.

18
I Then we use the previous result

(ab) mod m = (a mod m)(b mod m) mod m

I Hence
k−1
b n mod m = (b ak−1 2 mod m) . . . (b a1 2 mod m)(b a0 mod m) mod m

I Each of these factors can be computed from the previous one:

j j−1 j−1
b 2 mod m = (b 2 mod m)(b 2 mod m) mod m

19
Example
What is 367 mod 7?
I Apply the previous result:

3 mod 7 = 3
32 mod 7 = 9 mod 7 = 2
34 mod 7 = (32 mod 7)(32 mod 7) mod 7 = 4
38 mod 7 = (34 mod 7)(34 mod 7) mod 7 = 2
316 mod 7 = (38 mod 7)(38 mod 7) mod 7 = 4
332 mod 7 = (316 mod 7)(316 mod 7) mod 7 = 2
364 mod 7 = (332 mod 7)(332 mod 7) mod 7 = 4

I Since 67 = (1000011)2 = 26 + 2 + 1 = 64 + 2 + 1

367 mod 7 = (364 32 3 mod 7) = (4 · 2 · 3) mod 7 = 3

20
I In pseudocode:
procedure Modular Exponentiation(b : integer,
n = (an−1 . . . a1 a0 )2 , m : positive integers)
x := 1
power := b mod m
for j := 0, . . . , n − 1 do
if aj = 1 then
x := (x · power) mod m
end if
power := (power · power) mod m
end for
return x = b n mod m
end procedure

21
On optimal exponentiation

Given the task to calculate b 15 . With the binary approach we need to

calculate b 2 , b 4 = (b 2 )2 , b 8 = (b 4 )2 and finally b 15 = b · b 2 · b 4 · b 8 - 6
multiplications in total.
However, we can also calculate b 2 , b 3 = b 2 · b, b 6 = (b 3 )2 , b 9 = b 3 · b 6
and finally b 15 = b 6 · b 9 - 5 multiplications in total!
15 is the smallest exponent where the binary algorithm is not optimal.
Finding the optimal method in general is very hard3 .

3 https://en.wikipedia.org/wiki/Addition-chain_exponentiation
22