Learning Goals

• measure the performance of Ruby code
• determine why one instruction could be slower or faster than several instructions
• show how size of dataset can impact the performance of code
• benchmark various tasks (getting data from RAM, disk, API)

Agenda

• 2 min - Goals
• 20 min - Benchmarking Intro
• 30 min - Playing with Benchmark
• 15 min - Profiling Intro
• 20 min - Playing with Profiling
• 5 min - Wrap up

Materials

• Repo

Key Points

• Code is not free.
• All instructions are not created equally.

Tools

• Benchmark (built into Ruby)
• Profile (built into Ruby)
• Ruby-Prof (Gem)

Benchmark

• Used to measure and report time used to execute Ruby code
• part of Standard Library (must be required)
• set a variable to run code a certain number of times (x = 1_000_000)

Benchmark.bm

Benchmark.bm do |performance|
  performance.report('thing1') { 
    # your code here
  }

  performance.report('thing2') { 
    # your code here
  }
  ...
end

Benchmark.bmbm

The times may vary dramatically between running Benchmark.bm. In an attempt to standardize benchmark times, Benchmark.bmbm runs a "rehearsal", then runs GC.start, then runs the real thing.

Benchmark.bmbm do |performance|
  performance.report('thing1') { 
    # your code here
  }

  performance.report('thing2') { 
    # your code here
  }
  ...
end

Output

• Real time is the amount of time that a stopwatch or clock would measure between the start and end of process.
• User time is the time spent on running code that the user typed.
• System time is the time spent on running system instructions.
• Total time is User + System time.

Benchmark Exercises

• Loops
• Combining Strings
• Array Methods
• Destructive vs. Non-Destructive Methods
• Variable assignment
• Accessing data with attr_reader vs. instance variable
• Reading data from file vs. reading data from memory
• Conditional vs. Rescue

Profiling Code

• Profilers are used to determine which methods are called, how many times they are called, and how long each method is taking.
• Profilers are useful for finding bottlenecks in your code.

Ruby Profiler

ruby -rprofile filename.rb