How to benchmark Ruby code

Article originally published here : https://saaslit.com/blog/ruby/how-to-benchmark-ruby-code

What is benchmarking?

Benchmarking is the measure of performance of a piece of code. I personally find this area a little controversial. since something more readable, but that doesn't perform as well as a less performant piece of code.

But after playing with some Ruby methods, here are 3 ways that I tried to see if one way to solve a problem performs better than another.

Let's try a Ruby way to...

Today I'm going to try a while loop in Ruby. There are multiple available syntax, so I'm going to try 2 of them.

It will be mostly an example-based article. By googling around, I found 3 ways to benchmark a while loop. Let's go.

First, Benchmark.realtime

require 'benchmark'

time = Benchmark.realtime do
  x = 1
  loop do
    x += 1
    break if x > 100
  end
end
puts "loop took #{time} seconds."

time = Benchmark.realtime do
  x = 1
  while true
    x += 1
    break if x > 100
  end
end
puts "while took #{time} seconds."

Will result in:

loop took 8.14200029708445e-06 seconds.
while took 1.203499959956389e-05 seconds.

Ok, not super-readable report, but it seems that loop performs better.

Second, Benchmark.bm

require 'benchmark'

Benchmark.bm do |benchmark|
  benchmark.report("loop.") do
    x = 1
    loop do
      x += 1
      break if x > 100
    end
  end

  benchmark.report("while") do
    x = 1
    while true
      x += 1
      break if x > 100
    end
  end
end

Gives following results:

       user     system      total        real
loop.  0.000020   0.000006   0.000026 (  0.000022)
while  0.000003   0.000001   0.000004 (  0.000003)

"real" column gives the total time spent.

Yikes! It seems that now that the while version performs better.

Let's try once again with the same method.

       user     system      total        real
loop.  0.000014   0.000002   0.000016 (  0.000011)
while  0.000004   0.000000   0.000004 (  0.000004)

Ok it seems that 2 consecutive launch gives very different result. At least this time, the same method performs better.

Third, Benchmark.measure

require 'benchmark'

puts Benchmark.measure {
  50_000.times do
    x = 1
    loop do
      x += 1
      break if x > 100
    end
  end
}

puts Benchmark.measure {
  50_000.times do
    x = 1
    while true
      x += 1
      break if x > 100
    end
  end
}

It gives the following results :

  0.261000   0.003973   0.264973 (  0.266082)
  0.083610   0.000001   0.083611 (  0.083994)

The first line is the result about the .loop method, and the second one about the .while method.

The result between parenthesis is the aggregate (full time to compute the operation).

.while seems to be better here.

Summary

We have seen 3 ways to measure performance of a Ruby method, however, it's hard to conclude about anything, since I had different results here. I need to investigate, in order to be sure about what is going on.