Micro-benchmark for `value != nil` vs. `!value.nil?`

benchmark.rb

#!/usr/bin/env ruby

require 'benchmark'

Benchmark.bm do |b|
  n = 1_000_000

  value1 = 1
  value2 = nil

  b.report('value != nil') do
    n.times do
      value1 != nil
      value2 != nil
    end
  end

  b.report('!value.nil?') do
    n.times do
      !value1.nil?
      !value2.nil?
    end
  end
end

ruby-3.1.1:

       user     system      total        real
value != nil  0.376474   0.000000   0.376474 (  0.380029)
!value.nil?  0.074744   0.000000   0.074744 (  0.075391)

ruby-3.2.0 (no JIT):

       user     system      total        real
value != nil  0.266440   0.000851   0.267291 (  0.269028)
!value.nil?  0.053540   0.000000   0.053540 (  0.053805)

and if we also test nil != value, it's only slightly slower:

       user     system      total        real
value != nil  0.332498   0.000000   0.332498 (  0.334777)
nil != vaule  0.072327   0.000000   0.072327 (  0.072743)
!value.nil?  0.061445   0.000000   0.061445 (  0.061947)

Why?

@AndrewJGregory Faster due to it bit flipping?

ruby 2.7.6p219:

               user     system      total        real
value != nil  0.137024   0.000028   0.137052 (  0.137107)
!value.nil?  0.031404   0.000000   0.031404 (  0.031418)

This should be a rubocop cop, shouldn't it?

@andreavocado it actually is! At first I didn't believe it would have improved the code, which is why I wrote the benchmark.