Analysis of Algorithms

Analysis of Algorithms
Day 1
Objectives of the course

To introduce the concept of Analysis of Algorithms. To learn the various factors that affect the performance of an algorithm. To learn Code Tuning Techniques. To Analyze well known Algorithms. To introduce Intractable problems.
Copyright 2004, Infosys Technologies Ltd
ER/CORP/CRS/SE15/003 Version No: 2.0
References
1. Weiss M, W. (1993) Data Structures and Algorithm Analysis in C, Benjamin Cummings. 2. McConnell, S. (1993) Code complete, Microsoft Press 3. Alfred V Aho, John E Hopcraft, Jeffrey D Ullman, Design & Analysis of Computer Algorithms, Addison Wesley Publishing Company. 4. Ellis Horowitz, Sartaj Sahni, Sanguthevar, (1998)Fundamentals of Computer Algorithms, Galgotia Publications private limited, New Delhi. 5. Donald E Knuth (1938)The Art of Computer Programming, Fundamental Algorithms, Volume 1, Third Edition, Addison Wesley
Course Plan Day 1

Introduction to Analysis of algorithms
What is an Algorithm? Properties of an Algorithm Life cycle of an Algorithm
Code Tuning Techniques SQL Tuning Techniques
Course Plan (cont.) Day 2

Analyzing Algorithms
Basic Mathematical principles Order of magnitude Introduction to Asymptotic notations Best case Worst case Average case
Analysis of well known algorithms

Algorithm design techniques ( Brute force, Greedy, Divide & Conquer, Decrease & Conquer) Analysis of some well known algorithms
Course plan (cont.) Day 3

Algorithm Design techniques (contd.) (Dynamic Programming) Intractable problems
Deterministic Vs Non-Deterministic machines P Vs NP NP Complete NP Hard
Case Study
Plan for Day-1

Unit 1 - Introduction
Introduction to Analysis of Algorithms

Where does the word algorithm came from? King Algor of Castile ? Algiros + Arithmos meaning Painful + Numbers ? NO The etymology of the word Algorithm dates back to the 8th Century AD. The word Algorithm is derived from the name of the Persian author Abu Jafar Mohammad ibn Musa al Khowarizmi
Muhammad al-Khowarizmi, from a 1983 USSR commemorative stamp scanned by Donald Knuth Reference: ACM Trans - Algorithms
What is an Algorithm?
Finite set of instructions to accomplish a task. The properties of an algorithm are as follows:
Finiteness
Effectiveness
Algorithm
Definiteness
Output
Input
10
Pseudo Code
An algorithm is independent of any language or machine whereas a program is dependent on a language and machine. To fill the gap between these two, we need pseudo codes. Psuedo-code is a way to represent the step by step methods in finding the solution to the given problem. Example: Algorithm arrayMax (A,n) Input array A of n integers Output maximum element of A CurrentMax A[0] for I = 1 to n-1 do if A[i] > currentMax then currentMax A[i] return currentMax
11
Pseudo Code
An Algorithm should be independent of any programming language. Psuedo-code is a way to represent the step by step methods in finding the solution to the given problem. Example:
Algorithm arrayMax (A, n) Input array A of n integers Output maximum element of A currentMax for I A[0]
1 to n-1 do if A [i] > currentMax then currentMax A [i]
return currentMax
12
Life Cycle of an Algorithm

Design the Algorithm Write (Implementation of the Algorithm) Test the Algorithm Analyze the Algorithm
13
Life Cycle of an Algorithm (Contd)
14
Resources available in a computer
15
An algorithm when implemented, uses the computers primary memory and Central Processing Unit. Analyzing the amount of resources (time &/ space) needed for solving the problem). The Analysis is done at two stages:
Priori Analysis:
Analysis done before implementation.

Posteriori Analysis:
Analysis done after implementation.
16
Efficiency Measures
Performance of a solution
Most of the software problems do not have a single best solution. Then how do we judge these solutions? The solutions are chosen based on performance measures.
Performance Measures
Time Quality
Simplicity
17
Efficiency Measures (Contd)

Space Time Tradeoff
Example 1: Consider a personnel management product that an organization can purchase and use to maintain information about its employees. If employee details were to be stored in an array, the array would have to be declared large enough to be able to hold the maximum number of records the system was rated to handle. This would always take up a large amount of memory. With a linked list implementation on the other hand, there would be better utilization of memory. Which implementation would provide faster access to an employee with a given employee number? Which implementation would be easier to code? Which implementation would be easier to test?
18
Efficiency Measures (Contd )

Example 2: Think of a GUI drop-down list box that displays a list of employees whose names begin with a specified sequence of characters. If the employee database is on a different machine, then there are two options: Option a: fire a SQL and retrieve the relevant employee names each time the list is dropped down. Option b: keep the complete list of employees in memory and refer to it each time the list is dropped down. In your opinion which is the preferred option and why?
19
Efficiency Measures (Contd )

Example 3: Which of the following programs is more complicated? Design a computer program which produces an output 1, if the word is of length 3n (n=0,1,2,) and 0, otherwise. example: If the input is aabcef the output is 1 If the input is aabc then the output is 0 Design a computer program that sorts ( in Ascending order ) and outputs the result for any input sequence a1,a2,an of numbers, where n is any natural number. Hint: consider the RAM size required in both the programs
20
Summary of Unit-1
What is an Algorithm? Properties of an Algorithm. Life Cycle of an Algorithm Performance Measures
21
Unit 2 - Code Tuning Techniques
Code Tuning Techniques
Code Tuning refers to modifying the implementation of a specific design rather than modifying the design itself Related to how to code better How to review a given code with performance as a concern
23
CTT - Loops
Looping constructs in a program are executed many times. There are quite a few of code tuning techniques which can be applied to looping constructs in order to improve the performance of the code. Jamming of loops: Jamming of loops is an outcome of combining loops which operate over the same range of values.
for (k = 1 to n) { initialize T[k] }; for (m = 1 to n) { Max[m] = m * Max[m] }; for (k = 1 to n) { initialize T[k]; Max[k] = k * Max[k] };
Copyright 2004, Infosys Technologies Ltd 24
Before Jamming, 2n Loop Checks
After Jamming, n Loop Checks
CTT Loops (Contd)

Unswitching of Loops:
Unswitch loops that contain if tests, if the results of these tests do not change inside the loop
Let 10,000 employees of a company got loan in a bank. At the start of every month, the bank has to deduct the monthly installment of the loan from each employee if salary is credited in their account. for (i=0;i<noofemployees;i++){ If (salary_credited = YES) { if (salary_credited = YES) { for (i=0; i<noofemployees;i++){ ........ ........ ........ ........ } } }}
Move the if condition outside the loop

Copyright 2004, Infosys Technologies Ltd 25 ER/CORP/CRS/SE15/003 Version No: 2.0
CTT Loops (Contd)

Unrolling of Loops:
i = 1; while(i < num) { a( i ) = i; i = i + 1; }
Before unrolling, n loop checks
i = 1; while(i < num) { a( i ) = i; a( i + 1 ) = i + 1; i = i + 2; }
After unrolling, n/2 loop checks
26
CTT Loops (Contd)

Minimize work performed inside loops:
For (i = 1 to n/2) { Fprimes(i * i); Fprimes(i * i * i); }
n_2 = n/2; For (i = 1 to n_2) { m=i * i; Fprimes(m); Fprimes(m * i); }
Need to compute n/2 in every iteration is removed. Load of the loop is reduced.
27
CTT Loops (Contd)

Use of Sentinel Values:
In a character string in C language, the \0 is the sentinal
While (i< n) and (x< >a[i]) { i= i+1; }
a[n+1] = x; While (x< >a[i]) { i = i+1; }
One boundary check is reduced, as it is achieved by the sentinel

28
CTT Loops (Contd)

Reduce the strength of operations performed inside loops:
All operations which do not depend on loop variant may be moved outside loops and the operations are converted into cheaper ones. for (i = 1 to Num ) { commission ( i ) = i * Revenue * BaseCommission * Discount }
comm = Revenue * BaseCommission * Discount sum = comm for ( i = 1 to Num ) { commission ( i ) = sum sum = sum + comm }
29
CTT Logic
Order tests in case statements by frequency:
read (empNo) case Grade(empNo) of 1: {.} 2: {.} 3: {.} . 7: {.} endcase
read (empNo) case Grade(empNo) of 7: { .} 5: {.} 6: {.} . 1: {.} endcase
Every time, the more frequent, Grade 7 is encountered, the code executes faster
30
CTT Logic (Contd)

Stop testing when you know the result:
if(a < 10 ) and (b < 20) then { . }
if(a < 10) then { if(b < 20) then { . } }
While doing linear search come out as soon as the element is found.
31
CTT Data Transformations

Minimize array references:
If the same array element is repeatedly referred to inside a loop, then move it outside the loop
for (a=0; a < 5; a++) { for (b =0; b < 10; b++) { total[b] = total[b] * sum[a]; }
for (a=0; a < 5; a++) { sum_now = sum[a]; for (b =0; b < 10; b++) { total[b] = total[b] * sum_now; }
32
CTT Data Transformations (Contd)

Augment data structures with indexes:
For example we can add an index to the linear linked list data structure. This index helps in speeding up the search operation in a linked list which is other wise strictly linear.
In a character array, the length can be augmented at the start, so that finding the length of the string can be done easily without actually calculating it every time.
If a linked list is having 1000 elements, then searching the list on an average will cost 500 operations. If the same linked list is augmented with another small list of pointers each pointing to every 10th element in the original list, then the number of searches is reduced to an average of 55.
CTT Expressions
Use constants of the correct type:
float x; x = 5;
Convert 5 to 5.0 store into x
int i; i = 3.14;
Convert 3.14 to 3 store into i
Converting to a desired type is an overhead
34
CTT Expressions (Contd)

Precompute results:
y = log(x) / log(2) b = log(a) / log(2) instead have LOG2 = log(2) y = log(x) / LOG2 b = log(a) / LOG2
Four Function Calls
Three Function Calls
35

Exploit Algebraic Identities: Algebraic identities can be used to replace costlier operations by cheaper ones
< y, we can use the algebraic identity which says x < y only when x < y. So it is enough to check if x < y
in this case.
Whenever we need to find whether x
not (A or B) is cheaper than not A and not B
36

Strength Reduction in Expressions: Strength reduction refers to replacing costlier operations by cheaper ones. This can be achieved by:
Replacing multiplication with addition Replacing exponentiation with multiplication
Ax3 + Bx2 + Cx + D is better computed as ((Ax + B)x + C)x + D
37

Be Wary of System Routines: System Routines, like the math routines, provided accuracy which is more often wasted. If we do not need the level of accuracy as provided by the math routines then it makes sense for us to write the piece of code for the same. For example the math routine which computes log(x) provides the result in a floating point number whereas most of the times we might be interested only in the integral part.
38
SQL Query Tuning

Concatenation of different data types in SQL queries An SQL query involving concatenation of different data types takes more time to execute. Ex: SELECT * FROM dwtable2 WHERE empno||ename='1234name1234'; Instead, we can re write the above query as: SELECT * FROM dwtable2 WHERE empno=1234 AND ename='name1234';
39
SQL Query Tuning (Contd)

Position of table with fewer rows in the selectfrom query It is advisable to put the table that returns the fewest rows at the end of from list. Ex: Assume you have table1 with 10000000 records and table 2 with 1000 records. It is advisable to write the query as follows: Select count(*) from table1, table2 where rownum < 500;
40

Usage of Table Aliases If more than one table is used in a query, then it is advisable to use table aliases, as it would enhance the speed of parse phase of the query. Ex: SELECT idno, empno FROM emp1, emp2 WHERE emp1.idno = emp2.empno AND emp2.empno<=100; The same query can be written more efficiently as: SELECT idno,empno FROM emp1 e1, emp2 e2 WHERE e1.idno=e2.empno AND e2.empno<=100;
41

Usage of NOT or != operators Unless it is absolutely necessary, one should avoid the usage of NOT or != operator in SQL queries. Reason: Whenever these operators are used, a FTS (Full Table Scan) is done and will degrade the performance of the transactions.
42

Usage of Index/Indexes Index enables faster retrieval of data, but is an overhead when insertion, updation and deletion processes are involved.
43
Summary of Unit-2
Code Tuning Techniques for Loops Code Tuning Techniques for logic Code Tuning Techniques for Data Transformations Code Tuning Techniques for Expressions SQL Query Tuning
44
Summary of Day-1
45
Thank You!

Analysis of Algorithms

Enviado por

Dados do documento

Descrição original:

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Analysis of Algorithms

Enviado por

Direitos autorais:

Formatos disponíveis

Analysis of Algorithms

Objectives of the course

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Course Plan Day 1

Code Tuning Techniques SQL Tuning Techniques

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Course Plan (cont.) Day 2

Analysis of well known algorithms

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Course plan (cont.) Day 3

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Plan for Day-1

Code Tuning Techniques SQL Tuning Techniques

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Introduction to Analysis of Algorithms

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

1 to n-1 do if A [i] > currentMax then currentMax A [i]

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Life Cycle of an Algorithm

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Life Cycle of an Algorithm (Contd)

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Resources available in a computer

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Analysis done before implementation.

Analysis done after implementation.

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Efficiency Measures (Contd)

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Efficiency Measures (Contd )

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Efficiency Measures (Contd )

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Code Tuning Techniques

Copyright 2004, Infosys Technologies Ltd

ER/CORP/CRS/SE15/003 Version No: 2.0

Before Jamming, 2n Loop Checks

After Jamming, n Loop Checks