Introduction

• We will learn several ways to evaluate the performance of an algorithm or implementation.
  • Each has strengths and weaknesses.
  • But each has a place.
• For now, I would like to determine the performance of the string find function in the standard library.
• I suspect that this is linear (whatever that means)
  • Because sequential search is linear. (But this is a bad assumption)
  • But it is probably KMP or better.
• So what should I investigate?
  • The linear thing implies that as the data size grows, the time will grow at the same rate.
  • So if I just do a series of string finds and time them I should be in good shape.
• My plan:
  1. Generate a random text of a given length.
  2. Generate a random word of a given length.
  3. Search for the word in the text, finding all occurrences.
     • Time this.
• I need to see how tings run, so I might want to
  • Generate different size words and texts.
  • Run the experiment a number of times.
• I think I need the following tools
  • Learn how to time things inside a program.
    • I really don't want to consider the data generation times.
    • This means system level utilities like the time function in linux will not be helpful.
    • Furthermore, I would like to be somewhat portable.
  • Learn how to generate random data.
• The modern (c++11 or better) c++ libraries have new versions of libraries to assist with these tasks.
  • And we should look at them.
• The material for the next sections
  • Is from Gregorie, fifth edition.
  • Is only approached from a surface level.
    • I just want a tool set, I don't want to study c++, so we will not go in depth.
    • We don't have time.
    • Both of these topics are very deep,
      • Mathematically
      • And perhaps in other ways.
• You should use these ideas, if not this code to do your experiments (homework 1 and beyond).
  • So as we discuss them, if you don't understand something, ASK
  • And if you don't get the code, ASK
  • And if you don't understand the ideas, ASK