CSEE 3827: Fundamentals of Computer Systems,

Spring 2011

1. Number Representation

Prof. Martha Kim ([email protected])

Web: http://www.cs.columbia.edu/~martha/courses/3827/sp11/
Contents (H&H 1.3-1.4, 5.3)

• Digital Information Representation

• Decimal
• Hexadecimal
• BCD
• Terminology:
• Bit / Byte / Words
• Highest Order (most significant) Bit, Lowest Order (least significant) bit
• Negative Number Formats:
• Signed Magnitude
• 1’s Complement
• 2’s Complement
• Fractions via Binary
• Fixed Point
• Floating Point
2
Number systems: Base 10 (Decimal)

• 10 digits = {0,1,2,3,4,5,6,7,8,9}

• example: 4537.8 = (4537.8)

10

4 5 3 7 . 8
3 2 1 0 -1
x 10 x 10 x 10 x 10 x 10

4000 + 500 + 30 + 7 + .8 = 4537.8

Number systems: Base 2 (Binary)

• 2 digits = {0,1}

• example: 1011.1 = (1011.1)

2

1 0 1 1 . 1
3 2 1 0 -1
x 2 x 2 x 2 x 2 x 2

8 + 0 + 2 + 1 + .5 = (11.5) 10
Number systems: Base 8 (Octal)

• 8 digits = {0,1,2,3,4,5,6,7}

• example: (2365.2)
8

2 3 6 5 . 2
3 2 1 0 -1
x 8 x 8 x 8 x 8 x 8

1024 + 192 + 48 + 5 + .25 = (1269.25)10

Number systems: Base 16 (Hexadecimal)

• 16 digits = {0,1,2,3,4,5,6,7,8,9,A,B,C,D,E,F}

• example: (26BA) 16 [alternate notation for hex: 0x26BA]

2 6 B A
3 2 1 0
x 16 x 16 x 16 x 16

8192 + 1536 + 176 + 10 = (9914)10

Why Important: More concise than binary, but related (a power of 2)

Hexadecimal (or hex) is often used for addressing
Number ranges

• Map infinite numbers onto finite representation for a computer

• How many numbers can I represent with ...

... 5 digits in decimal?

5
10 possible values

... 8 binary digits? 8

2 possible values

... 4 hexadecimal digits? 4

16 possible values
Computer from Digital Perspective

• Information: just sequences of binary (0’s and 1’s)

• True = 1, False = 0

• Numbers: converted into binary form when “viewed” by computer

• e.g., 19 = 10011 (16 (1) + 8 (0) + 4 (0) + 2 (1) + 1 (1)) in binary

• Characters: assigned a specific numerical value (ASCII standard)

• e.g., ‘A’ = 65 = 1000001, ‘a’ = 97 = 1100001

• Text is a sequence of characters:

• “Hi there” = 72, 105, 32, 116, 104, 101, 114, 101

= 1001000, 1101001, ...

Terminology: Bit, Byte, Word

• bit = a binary digit e.g., 1 or 0

• byte = 8 bits e.g., 01100100

• word = a group of bits that is architecture dependent

(the number of bits that an architecture can process at once)

a 16-bit word = 2 bytes e.g., 1001110111000101

a 32-bit word = 4 bytes e.g., 100111011100010101110111000101

OBSERVATION: computers have bounds on how much input they can

handle at once limits on the sizes of numbers they can deal with
Terminology: MSB, LSB

• Bit at the left is highest order, or most significant bit (MSB)

• Bit at the right is lowest order, or least significant bit (LSB)

MSB LSB

e.g., 1001110111000101

• Common reference notation for k-bit value: bk-1bk-2bk-3...b1b0

Unsigned numbers a.k.a. Binary Coded Decimal (BCD)

Binary numbers represent only non-negative (positive or 0) values

BCD where wordsize=3:

value BCD
0 000
1 001
2 010
3 011
4 100
5 101
6 110
7 111
Addition of binary (unsigned numbers)
Like decimal addition, except:
1+1 = 0 with a carry of 1,
1+1+1 = 1 with carry of 1
e.g., wordsize = 5, add 11110 and 10101 (30 + 21)

Overflow 1 1 1
when result cannot fit 11110 30
within the wordsize + 21
constraint 10101
1 0 0 1 1 19
e.g., the “correct” answer
110011 requires 6 bits: cannot
be represented with only 5
bits in unsigned representation
What about negative numbers?

Given a fixed wordsize how do you represent both positive and negative
numbers?

Have certain bit combinations represent negative numbers

• e.g., Signed Magnitude
• highest order bit (bk-1) indicates sign: 0 = positive, 1 = negative
• remaining bits indicate magnitude
• e.g., 0011 = 3
• e.g., 1011 = -3
• e.g., 1000 = 0000 = 0
• Positive #’s have same form in both signed magnitude and unsigned
• Easy for humans to interpret, but not easiest form for computers to do
addition/subtraction operations
 Negative Numbers: 1’s Complement Representation

• Non-negative #’s have same representation as unsigned (and signed-mag)

• To negate a #, flip all bits (not just highest-order as in signed-mag)
• e.g., wordsize = 4
• 0010 = 2
• 1101 = -2

Suppose wordsize is 8, what is the value of 11101011 when it represents a # in 1’s

Complement representation?
Let x=11101011; Know X is negative because MSB=1.
Negate X by flipping all bits: -X = 00010100
-X = 20, so X = -20

Note: in 1’s complement, there are two ways to represent 0: all 0s and all 1s
 Negative Numbers: 2’s Complement Representation

• Non-negative #’s have same representation as unsigned (and signed-mag)

• To negate a #, flip all bits and add 1
• e.g., wordsize = 4
• 0010 = 2, so 1101 + 1 = 1110 = -2
• 0110 = 6, so 1001 + 1 = 1010 = -6
• 1010 = -6, so 0101 + 1 = 0110 = 6 (works in both directions)
• 0000 = 0, so 1111 + 1 = 0000 = 0 (0 is unique in 2’s complement)

Note: negation works both ways in all cases except 1 followed by all 0s (e.g., 1000).
for wordsize=k, the value is -2k-1 (e.g., k=4, value is -8)
Note: the positive value of 2k-1 is not expressible
Number encoding summary

BCD Sign&Mag. 1s Comp. 2s Comp.

000 0 +0 +0 +0
001 1 +1 +1 +1
010 2 +2 +2 +2
011 3 +3 +3 +3
100 4 0 -3 -4
101 5 -1 -2 -3
110 6 -2 -1 -2
111 7 -3 0 -1
8 values 7 values, 7 values, 8 values,
2 zeroes 2 zeroes 1 zero
k-bit Words & Ranges of various formats

• Given a k-bit word, what range of numbers can be represented as:

• unsigned: 0 to 2k - 1 (e.g., k=8, 0 to 255)

• signed mag: -2k-1 + 1 to 2k-1 - 1 (e.g., k=8, -127 to 127 [2 vals for 0])

• 1’s complement: same as signed mag (but negative numbers are

represented differently)

• 2’s complement: -2k-1 to 2k-1 - 1 (e.g., k=8, -128 to 127 [1 val for 0])
Getting representation

Given an 8-bit wordsize, what is the value of 10001011?

What do you mean, Unsigned, Signed Magnitude, 1’s complement or 2’s

complement?

• Unsigned: 128 + 8 + 2 + 1 = 139

• Signed Mag: -1 * (8 + 2 + 1) = -11

• 1’s Complement: the negation of 01110100 = -116

• 2’s Complement: 1’s complement + 1 = -117

Representation v. Operation

• We have discussed various representations for expressing integers

• unsigned, signed magnitude, 1’s-complement, 2’s-complement
• There are also bit-oriented operations that go by the same names
• 1’s-complement: flip all bits
• 2’s-complement: flip all bits and add 1
• Operation can be performed on a number, regardless of representation
• e.g., let 10111 be a number in signed-magnitude form (value is -7)
• 2’s complement (operation) of 10111 = 01001 (value is 9 in signed-mag form)

• Observe:
• 2’s-complement operation negates a number when in 2’s-complement
representation

• 1’s-complement operation negates a number when in 1’s-complement

representation
Automating Subtraction
Why are we interested in 2‘s-complement when it seems so less intuitive?

Much easier to automate subtraction (i.e., add #’s of opposite sign)

• Just negate subtrahend (bottom # in subtract) and add

• e.g, wordsize 6, perform 14 - 21 using signed magnitude representation

001110 001110
- +
101011
negate subtrahend (2’s complement op)
010101
111001

X = 111001, -X = 000111 = 7, X=-7

Why 2’s-complement subtraction works (basic idea)
0
• Think of a pinwheel, here is BCD representation
7 000 1
111 001
• Addition operation of X+Y interprets Y as
6 110 010 2
BCD and shifts Y slots clockwise from X to
give the sum 101 011
5 100 3

• Now change the representation to 2’s complement 4

0
• X+Y still shifts bits (Y in BCD) slots clockwise
-1 000 1
111 001
• e.g., 2-3 = 010+101 = 010 shifted 5 slots
clockwise = 111 = -1 -2 110 010 2

Another nice feature of 2s complement representation = easy to 101 011

detect overflow. More on that later. Remainder of the course:
-3 100 3
unsigned or 2s complement.
-4
What about numbers with fractions?

• Two common notations

• Fixed-point (the binary point is fixed)

• Floating-point (the binary point floats to the right of the most

significant 1)
Fixed-Point Notation

• Fixed-point representation of 6.75 using 4 integer bits and 4 fraction bits:

01101100
0110.1100
2 1 -1 -2
2 + 2 + 2 + 2 = 6.75

• The binary point is not a part of the representation but is implied.

• The number of integer and fraction bits must be agreed upon by those
generating and those reading the number.
Floating-Point Notation

• The binary point floats to the right of the most significant 1.

• Similar to decimal scientific notation.

• For example, write 27310 in scientific notation: 2

273 = 2.73 × 10

• In general, a number is written in scientific notation as: ± M × BE

• Where, M = mantissa, B = base, E=exponent

Need a bigger range?

• Change the encoding.

• Floating point (used to represent very large numbers in a compact way)

5 exponent
• A lot like scientific notation: 5.4 x 10
mantissa

• Except that it is binary: 1011

1001 x 2
What about negative numbers?

• Change the encoding.

• Sign and magnitude

• Ones compliment

• Twos compliment
Sign and magnitude

• Most significant bit is sign

• Rest of bits are magnitude

0110 = (6)10 1110 = (-6) 10

• Two representations of zero

0000 = (0) 1000 = (-0)

10 10
Ones compliment

• Compliment bits in positive value to create negative value

• Most significant bit still a sign bit

0110 = (6)10 1001 = (-6) 10

• Two representations of zero

0000 = (0) 1111 = (-0)

10 10
Twos compliment

• Compliment bits in positive value and add 1 to create negative value

• Most significant bit still a sign bit

0110 = (6)10 1001 + 1 = 1010 = (-6) 10

• One representation of zero

0000 = (0) 1000 = (-8) 1111 = (-1)

10 10 10
• One more negative number than positive

MIN: 1000 = (-8) MAX: 0111 = (7)

10 10
How about letters?

• Change the encoding.

Gray code
Binary numeric encoding where successive numbers differ by only 1 bit

value BCD # bit flips Gray # bit flips

0 000 3 000 1

1 001 1 001 1

2 010 2 011 1

3 011 1 010 1

4 100 3 110 1

5 101 1 111 1

6 110 2 101 1

7 111 1 100 1
How about letters?

• Change the encoding.