CH 8 Stallings

William Stallings Computer
Organization and Architecture
Chapter 8
Computer Arithmetic
Arithmetic & Logic Unit
 Does the calculations (arithmetic & logical)

 Handles integers
 May handle floating point (real) numbers
 May be separate FPU (maths co-processor)
ALU Inputs and Outputs
Control unit provides signal

to control the followings:
• Operation of ALU
• Movement of data into and
out Of ALU
Flag (register) is used to check

Whether the result stored in Register
exceeds the length of the Register of
not. For example:
Stored inputs Stored output • Flag (overflow flag) is set into 1 if
results the result of computation exceeds
the length of the Register
Integer Representation
 Only have 0 & 1 to represent everything

 Positive numbers stored in binary
 e.g. 41=00101001
 No minus sign • If we are limited to non-negative
numbers then representation is
 No period straightforward.
 Sign-Magnitude
 Two’s compliment However, in real life, we need
to handle minus numbers and
floating values.
Sign-Magnitude
 Left most bit is sign bit • Remaining bits are magnitude
 0 means positive
 1 means negative
 +18 = 00010010
 -18 = 10010010
 Problems 1. Need to consider both sign and
magnitude in arithmetic
Problem of Sign-Magnitude 2. Also two representation of
representation: zero (+0 and -0)
+0 = 00000000
-0= 10000000
Two’s Compliment Example
Two’s Compliment
 +3 = 00000011
 +2 = 00000010
 +1 = 00000001
 +0 = 00000000
 -1 = 11111111
 -2 = 11111110
 -3 = 11111101
Benefits of 2’s Complement
 One representation of zero

 Arithmetic works easily (see later)
 Negating is fairly easy
 3 = 00000011 (HW implementation)
 Boolean complement gives 11111100
 Add 1 to LSB 11111101
Geometric Depiction of Twos
Complement Integers (4-bit number)
Range of Numbers
 8 bit 2s compliment
 +127 = 01111111 = 27 -1
 -128 = 10000000 = -27
 16 bit 2s compliment
 +32767 = 011111111 11111111 = 215 - 1
 -32768 = 100000000 00000000 = -215
Conversion Between Lengths
 Positive number pack with leading zeros

 +18 = 00010010
 +18 = 00000000 00010010
 Negative numbers pack with leading ones
 -18 = 10010010
 -18 = 11111111 10010010
 i.e. pack with MSB (sign bit)
Addition and Subtraction
 Normal binary addition

 Monitor sign bit for overflow
 Take twos compliment of substahend and add

to minuend
 i.e. a - b = a + (-b)
 So we only need addition and complement

circuits
Hardware for Addition and
Subtraction
Multiplication
 Complex
 Work out partial product for each digit
 Take care with place value (column)
 Add partial products
Multiplication Example (unsigned)
 1011 Multiplicand (11 dec)

 x 1101 Multiplier (13 dec)
 1011 Partial products
 0000 Note: if multiplier bit is 1 copy
 1011 multiplicand (place value)
 1011 otherwise zero
 10001111 Product (143 dec)
 Note: need double length result
Unsigned Binary Multiplication
Hardware Implementation
1011 = M (register)
x 1101 = Q (register)
1-bit reg
Third reg A: initially 0

1011 = M (register)
Execution of Example x 1101 = Q (register)
(Product in A, Q)
Flowchart for Unsigned Binary
Multiplication
Multiplying Negative Numbers
 This does not work!

 Solution 1
 Convert to positive if required
 Multiply as above
 If signs were different, negate answer
 Solution 2
 Booth’s algorithm
Home Task
Booth’s Algorithm
Example of Booth’s Algorithm
Division
 More complex than multiplication

 Negative numbers are really bad!
 Based on long division
Division of Unsigned Binary
Integers
00001101 Quotient
Divisor 1011 10010011 Dividend
1011
Partial 001110
1011
Remainders
001111
1011
Remainder
100
Integers
Integers
Real Numbers
 Numbers with fractions

 Could be done in pure binary
 1001.1010 = 24 + 20 +2-1 + 2-3 =9.625
 Where is the binary point?
 Fixed?
 Very limited
 Moving?
 How do you show where it is?
Floating Point
Sign bit
Biased Significand or Mantissa

Exponent
 +/- .significand x 2exponent

 Misnomer
 Point is actually fixed between sign bit and body
of mantissa
 Exponent indicates place value (point position)
Floating Point Examples
Signs for Floating Point
 Mantissa is stored in 2s compliment

 Exponent is in excess or biased notation
 e.g. Excess (bias) 128 means
 8 bit exponent field
 Pure value range 0-255
 Subtract 128 to get correct value
 Range -128 to +127
Normalization
 FP numbers are usually normalized

 i.e. exponent is adjusted so that leading bit
(MSB) of mantissa is 1
 Since it is always 1 there is no need to store it
 (c.f. Scientific notation where numbers are
normalized to give a single digit before the
decimal point
 e.g. 3.123 x 103)
FP Ranges
 For a 32 bit number

 8 bit exponent
 +/- 2256  1.5 x 1077
 Accuracy
 The effect of changing lsb of mantissa
 23 bit mantissa 2-23  1.2 x 10-7
 About 6 decimal places
Expressible Numbers
IEEE 754
 Standard for floating point storage

 32 and 64 bit standards
 8 and 11 bit exponent respectively
 Extended formats (both mantissa and exponent)
for intermediate results
FP Arithmetic +/-
 Check for zeros

 Align significands (adjusting exponents)
 Add or subtract significands
 Normalize result
FP Arithmetic x/
 Check for zero

 Add/subtract exponents
 Multiply/divide significands (watch sign)
 Normalize
 Round
 All intermediate results should be in double
length storage
Floating
Point
Multiplication
Floating
Point
Division
Required Reading
 Stallings Chapter 8
 IEEE 754 on IEEE Web site

CH 8 Stallings

Uploaded by

Copyright:

Available Formats

CH 8 Stallings

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CH 8 Stallings

Uploaded by

Copyright:

Available Formats

William Stallings Computer

Organization and Architecture

 Does the calculations (arithmetic & logical)

Control unit provides signal

Flag (register) is used to check

 Only have 0 & 1 to represent everything

 Left most bit is sign bit • Remaining bits are magnitude

 One representation of zero

 Positive number pack with leading zeros

 Normal binary addition

 Take twos compliment of substahend and add

 So we only need addition and complement

 1011 Multiplicand (11 dec)

Third reg A: initially 0

 This does not work!

 More complex than multiplication

 Numbers with fractions

Biased Significand or Mantissa

 +/- .significand x 2exponent

 Mantissa is stored in 2s compliment

 FP numbers are usually normalized

 For a 32 bit number

 Standard for floating point storage

 Check for zeros

 Check for zero

You might also like