loop_unroll.rb

# frozen_string_literal: true

require "benchmark/ips"

array = Array.new(1_000_000) { rand(100) }.freeze

def sum_with_loop(array)
  sum = 0
  array.each do |num|
    sum += num
  end
  sum
end

def sum_with_unrolled_loop_4_alt(array)
  sum = 0
  i = 0
  len = array.length
  while i <= len - 4
    sum += array.values_at(i, i + 1, i + 2, i + 3).sum
    i += 4
  end
  while i < len
    sum += array[i]
    i += 1
  end
  sum
end

def sum_with_unrolled_loop_4(array)
  sum = 0
  i = 0
  len = array.length
  while i <= len - 4
    sum += array[i] + array[i + 1] + array[i + 2] + array[i + 3]
    i += 4
  end
  while i < len
    sum += array[i]
    i += 1
  end
  sum
end

def sum_with_unrolled_loop_8(array)
  sum = 0
  i = 0
  len = array.length
  while i <= len - 8
    sum += array[i] + array[i + 1] + array[i + 2] + array[i + 3] + array[i + 4] + array[i + 5] + array[i + 6] + array[i + 7]
    i += 8
  end
  while i < len
    sum += array[i]
    i += 1
  end
  sum
end

def sum_with_unrolled_loop_16(array)
  sum = 0
  i = 0
  len = array.length
  while i <= len - 16
    sum += array[i] + array[i + 1] + array[i + 2] + array[i + 3] + array[i + 4] + array[i + 5] + array[i + 6] + array[i + 7] +
            array[i + 8] + array[i + 9] + array[i + 10] + array[i + 11] + array[i + 12] + array[i + 13] + array[i + 14] + array[i + 15]
    i += 16
  end
  while i < len
    sum += array[i]
    i += 1
  end
  sum
end

Benchmark.ips do |x|
  x.report("Standard Loop") { sum_with_loop(array) }
  x.report("Unrolled Loop x4") { sum_with_unrolled_loop_4(array) }
  # x.report("Unrolled Loop x4-alt") { sum_with_unrolled_loop_4_alt(array) }
  x.report("Unrolled Loop x8") { sum_with_unrolled_loop_8(array) }
  x.report("Unrolled Loop x16") { sum_with_unrolled_loop_16(array) }
  x.compare!
end

result.txt

🍔 ruby loop_unroll.rb
ruby 3.4.5 (2025-07-16 revision 20cda200d3) +PRISM [arm64-darwin24]
Warming up --------------------------------------
       Standard Loop     3.000 i/100ms
    Unrolled Loop x4     6.000 i/100ms
    Unrolled Loop x8     7.000 i/100ms
   Unrolled Loop x16     8.000 i/100ms
Calculating -------------------------------------
       Standard Loop     36.917 (± 2.7%) i/s   (27.09 ms/i) -    186.000 in   5.040707s
    Unrolled Loop x4     67.065 (± 1.5%) i/s   (14.91 ms/i) -    336.000 in   5.012591s
    Unrolled Loop x8     73.777 (± 2.7%) i/s   (13.55 ms/i) -    371.000 in   5.031966s
   Unrolled Loop x16     81.351 (± 2.5%) i/s   (12.29 ms/i) -    408.000 in   5.018836s

Comparison:
   Unrolled Loop x16:       81.4 i/s
    Unrolled Loop x8:       73.8 i/s - 1.10x  slower
    Unrolled Loop x4:       67.1 i/s - 1.21x  slower
       Standard Loop:       36.9 i/s - 2.20x  slower