Computer Fundamentals:

Number Systems

Dr Robert Harle
Today's Topics
 The significance of the bit and powers of 2
 Data quantities (B, kB, MB, GB, etc)
 Number systems (decimal, binary, octal , hexadecimal)
 Representing negative numbers (sign-magnitude, 1’s
complement, 2’s complement)
 Binary addition (carries, overflows)
 Binary subtraction
So...
 What is a bit?
The Significance of the Bit
 A bit (Binary digIT) is merely 0 or 1
 It is a unit of information since you cannot communicate
with anything less than two states
 The use of binary encoding dates back to the 1600s with
Jacquard's loom, which created textiles using card
templates with holes that allowed needles through
Bits and Computers
 The nice thing about a bit is that, with only two states, it is
easy to embody in physical machinery
 Each bit is simply a switch and computers moved from
vacuum tubes to transistors for this

e-
Decimal Number System
 Most computers count in binary, which we can easily
understand from the decimal so ingrained in us

35462
Decimal Number System
 Most computers count in binary, which we can easily
understand from the decimal so ingrained in us

2x100

35462
Decimal Number System
 Most computers count in binary, which we can easily
understand from the decimal so ingrained in us

6x101+2x100

35462
Decimal Number System
 Most computers count in binary, which we can easily
understand from the decimal so ingrained in us

4x102+ 6x101+2x100

35462
Decimal Number System
 Most computers count in binary, which we can easily
understand from the decimal so ingrained in us

5x103+4x102+ 6x101+2x100

35462
Decimal Number System
 Most computers count in binary, which we can easily
understand from the decimal so ingrained in us

3x104+ 5x103+4x102+ 6x101+2x100

35462
Binary
 Binary is exactly the same, only instead of ten
digits/states (0 to 9) we have just two, so the base
becomes 2:

1x24+ 0x23+ 1x22+ 1x21+ 0x20

10110
Binary
 Binary is exactly the same, only instead of ten
digits/states (0 to 9) we have just two, so the base
becomes 2:

1x24+ 0x23+ 1x22+ 1x21+ 0x20

10110b = 22d
Binary
 Binary is exactly the same, only instead of ten
digits/states (0 to 9) we have just two, so the base
becomes 2:

1x24+ 0x23+ 1x22+ 1x21+ 0x20

10110b = 22d

Most Significant Least Significant

Bit (MSB) Bit (LSB)
Works for Fractional Numbers too...

3x101+ 5x100+ 4x10-1+6x10-2+2x10-3

35.462
Works for Fractional Numbers too...

3x101+ 5x100+ 4x10-1+6x10-2+2x10-3

35.462

1x21+ 0x20+ 1x2-1+ 1x2-2+ 0x2-3

10.110b = 2.75d
Check

11.011b
Check

1x21+ 1x20+ 0x2-1+ 1x2-2+ 1x2-3

11.011b = 3.375d
Representable Numbers
 With d decimal digits, we can represent 10d different
values, usually the numbers 0 to (10d-1) inclusive
 In binary with n bits this becomes 2n values, usually
the range 0 to (2n-1)

 Computers usually assign a set number of bits

(physical switches) to an instance of a type.
 An integer is often 32 bits, so can represent
positive integers from 0 to 4,294,967,295 incl.
 Or a range of negative and positive integers...
Other Common Bases
 Higher bases make for shorter numbers that are easier for
humans to manipulate. e.g.
6654733d=11001011000101100001101b
 We traditionally choose powers-of-2 bases because this
corresponds to whole chunks of binary
Other Common Bases
 Higher bases make for shorter numbers that are easier for
humans to manipulate. e.g.
6654733d=11001011000101100001101b
 We traditionally choose powers-of-2 bases because this
corresponds to whole chunks of binary
 Octal is base-8 (8=23 digits, which means 3 bits per digit)
 6654733d=011-001-011-000-101-100-001-101b= 31305415o
Other Common Bases
 Higher bases make for shorter numbers that are easier for
humans to manipulate. e.g.
6654733d=11001011000101100001101b
 We traditionally choose powers-of-2 bases because this
corresponds to whole chunks of binary
 Octal is base-8 (8=23 digits, which means 3 bits per digit)
 6654733d=011-001-011-000-101-100-001-101b= 31305415o
 Hexadecimal is base-16 (16=24 digits so 4 bits per digit)
 Our ten decimal digits aren't enough, so we add 6 new
ones: 0,1,2,3,4,5,6,7,8,9,A,B,C,D,E,F
 6654733d=0110-0101-1000-1011-0000-1101 b=658B0Dh
Other Common Bases
 Higher bases make for shorter numbers that are easier for
humans to manipulate. e.g.
6654733d=11001011000101100001101b
 We traditionally choose powers-of-2 bases because this
corresponds to whole chunks of binary
 Octal is base-8 (8=23 digits, which means 3 bits per digit)
 6654733d=011-001-011-000-101-100-001-101b= 31305415o
 Hexadecimal is base-16 (16=24 digits so 4 bits per digit)
 Our ten decimal digits aren't enough, so we add 6 new
ones: 0,1,2,3,4,5,6,7,8,9,A,B,C,D,E,F
 6654733d=0110-0101-1000-1011-0000-1101 b=658B0Dh
 Because we constantly slip between binary and hex, we have
a special marker for it
 Prefix with '0x' (zero-x). So 0x658B0D= 6654733d, 0x123=291d
Bytes
 A byte was traditionally the number of bits needed to
store a character of text
 A de-facto standard of 8 bits has now emerged
 256 values
 0 to 255 incl.
 Two hex digits to describe
 0x00=0, 0xFF=255
 Check: what does 0xBD represent?
Bytes
 A byte was traditionally the number of bits needed to
store a character of text
 A de-facto standard of 8 bits has now emerged
 256 values
 0 to 255 incl.
 Two hex digits to describe
 0x00=0, 0xFF=255
 Check: what does 0xBD represent?
 B → 11 or 1011
 D → 13 or 1101
 Result is 11x161+13x160 = 189 or 10111101
Larger Units
 Strictly the SI units since 1998 are:
 Kibibyte (KiB)
 1024 bytes (closest power of 2 to 1000)
 Mebibyte (MiB)
 1,048,576 bytes
 Gibibyte (GiB)
 1,073,741,824 bytes
Larger Units
 Strictly the SI units since 1998 are:
 Kibibyte (KiB)
 1024 bytes (closest power of 2 to 1000)
 Mebibyte (MiB)
 1,048,576 bytes
 Gibibyte (GiB)
 1,073,741,824 bytes
 but these haven't really caught on so we tend to still use the SI
Kilobyte, Megabyte, Gigabyte. This leads to lots of confusion
since technically these are multiples of 1,000.
The Problem with Ten
Unsigned Integer Addition
 Addition of unsigned integers works the same way as
addition of decimal (only simpler!)
 0+0=0
 0+1=1
 1+0=1
 1 + 1 = 0, carry 1
 Only issue is that computers have fixed sized types so we
can't go on adding forever...

001 111
+ 001 + 001

Carry flag: Carry flag:

Modulo or Clock Arithmetic

111 000
000
110 001 001
010
101 010 011
100 011
100
101
110
111

 Overflow takes us across the dotted

boundary
 So 7+1=0 (overflow)
 We say this is (7+1) mod 8
Negative Numbers
 All of this skipped over the need to represent
negatives.
 The naïve choice is to use the MSB to indicate +/-
 1 in the MSB → negative
 0 in the MSB → positive

-7 = 1111

Normal positive
Negative representation of 7

 This is the sign-magnitude technique

Difficulties with Sign-Magnitude
 Has a representation of minus zero (10002=-0) so
wastes one of our 2n labels
 Addition/subtraction circuitry must be designed
from scratch

1101
+ 0001
1110

Our unsigned addition alg.

Difficulties with Sign-Magnitude
 Has a representation of minus zero (10002=-0) so
wastes one of our 2n labels
 Addition/subtraction circuitry must be designed
from scratch

1101 +13
+ 0001 +1
1110 +14

Unsigned
interpretation

Our unsigned addition alg.

Difficulties with Sign-Magnitude
 Has a representation of minus zero (10002=-0) so
wastes one of our 2n labels
 Addition/subtraction circuitry must be designed
from scratch

-5 1101 +13
+1 + 0001 +1
-6 1110 +14

Sign-mag Unsigned
interpretation interpretation

Our unsigned addition alg.

Alternatively...
111 000 000
001
110 001
010
101 010 011
100
100 011 101
110
111

 Gives us two discontinuities and a

reversal of direction using normal
addition circuitry!!
Ones' Complement
 The negative is the positive with all the bits flipped
 7 → 0111 so -7 → 1000
 Still the MSB is the sign
 One discontinuity but still -0 :-(

111 000 000

001
110 001
010
101 010 011
100
100 011 101
110
111
Two's Complement
 The negative is the positive with all the bits flipped
and 1 added (the same procedure for the inverse)
 7 → 0111 so -7 → 1000+0001 → 1001
 Still the MSB is the sign
 One discontinuity and proper ordering

111 000 000

001
110 001
010
101 010 011
100
100 011 101
110
111
Two's Complement

 Positive to negative: Invert all the bits and add 1

0101 (+5) → 1010 → 1011 (-5)

 Negative to positive: Same procedure!!

1011 (-5) → 0100 → 0101 (+5)
Signed Addition
 ...it just works with our addition algorithm!

1101 +13
+0001 +1
1110 +14

Unsigned

Our unsigned addition alg.

 Signed Addition
 ...it just works with our addition algorithm!

-3 1101 +13
+1 +0001 +1
-2 1110 +14

2's-comp Unsigned

Our unsigned addition alg.

 Signed Addition
 ...it just works with our addition algorithm!

-3 1101 +13
+1 +0001 +1
-2 1110 +14

2's-comp Unsigned

Our unsigned addition alg.

Signed Addition
 So we can use the same circuitry for unsigned and 2s-
complement addition :-)
 Well, almost.

0100 +4
Carry flag: 0
+0100 +4
1000 +8

Unsigned
Signed Addition
 So we can use the same circuitry for unsigned and 2s-
complement addition :-)
 Well, almost.

+4 0100 +4
Carry flag: 0
+4 +0100 +4
-8 1000 +8

2's-comp Unsigned

 The problem is our MSB is now signifying the sign and our carry
should really be testing the bit to its right :-(
Signed Addition
 So we can use the same circuitry for unsigned and 2s-
complement addition :-)
 Well, almost.

+4 0100 +4
Carry flag: 0
+4 +0100 +4 Overflow: 1
-8 1000 +8

2's-comp Unsigned

 The problem is our MSB is now signifying the sign and our carry
should really be testing the bit to its right :-(
 So we introduce an overflow flag that indicates this problem
Integer subtraction
 Could implement the “borrowing”
algorithm you probably learnt in school
 But why bother? We can just add the
2's complement instead.

0100 0100
- 0011 → +1101
0001
Flags Summary
 When adding/subtracting
 Carry flag → overflow for unsigned integer
 Overflow flag → overflow for signed integer

 The CPU does not care whether it's

handling signed or unsigned integers
 Down to our compilers/programs to
interpret the result
Fractional Numbers
 Scientific apps rarely survive on integers alone, but
representing fractional parts efficiently is
complicated.
 Option one: fixed point
 Set the point at a known location. Anything to
the left represents the integer part; anything to
the right the fractional part
 But where do we set it??
 Option two: floating point
 Let the point 'float' to give more capacity on its
left or right as needed
 Much more efficient, but harder to work with
 Very important: more in Numerical Methods
course