Simple and delicious.

Today I ran the following script with Ruby 1.8 & Ruby 1.9 to compare their performances:

def bench
  start = Time.now
  1000000.times do
    yield
  end
  puts Time.now - start
end

puts "Test 1: do things"
bench {
  "yeho!12".next
  rand(100)
  i ||= 1
  i = i + 1
}

puts "Test 2: \"stuff\""
bench {
  "stuff"
}

puts "Test 3: 'stuff'"
bench {
  'stuff'
}

puts "Test 4: :stuff"
bench {
  :stuff
}

Ruby 1.9 performances are promising:

So Ruby 1.9 is 3 to 5 times faster than Ruby 1.8 to run simple operations. I then checked with a small Rails app.

Once I got rubygem installed for Ruby 1.9, the gems I needed installed for Ruby 1.9, the plug-ins I use patched for Ruby 1.9, and my ruby code patched for Ruby 1.9, – yes, it was painful! – I fired: time spec spec

Ruby 1.8

$> time spec spec
............................................

Finished in 0.594813 seconds

44 examples, 0 failures
spec spec  2.49s user 0.79s system 93% cpu 3.522 total

Ruby 1.9

$> time spec spec
............................................

Finished in 0.625589223 seconds

44 examples, 0 failures
spec spec  8.74s user 0.32s system 93% cpu 9.648 total

Grrrr. Ruby 1.8 & 1.9 both pass the specs in ~0.60 second but Ruby 1.9 takes 8.74 seconds in total vs 2.49 seconds for Ruby 1.8. The same behavior occurs when running a Webrick server via script/server: Ruby 1.9 is 2 times slower than Ruby 1.8 to boot up the server and it handles the requests just as fast as Ruby 1.8.

Any Ruby guru to explain such deceiving results?