Crystal BitArray methods with benchmarks of different #fill implementations

bit_array.cr

require "benchmark"
require "bit_array"
require "intrinsics"

struct BitArray
  def self.new(size, & : Int32 -> Bool)
    instance = BitArray.new(size)
    (0...size).each do |i|
      instance[i] = yield i
    end
    instance
  end

  def fill_override(value : Bool) : self
    @bits.to_slice(malloc_size).fill (value ? ~0u32 : 0u32)
    self
  end

  def fill_override_intrinsics(value : Bool) : self
    Intrinsics.memset(@bits, (value ? ~0u8 : 0u8), bytesize, false)
    self
  end

  def fill_override_intrinsics_bitwise(value : Bool) : self
    Intrinsics.memset(@bits, (value.to_unsafe << 8), bytesize, false)
    self
  end
end

RANDOM = Random.new

ba = BitArray.new(12) { |i| RANDOM.next_bool }
pp ba

puts ""

guide = BitArray.new(1_000) { |i| RANDOM.next_bool }
start = BitArray.new(1_000) { |i| RANDOM.next_bool }
Benchmark.ips do |x|
  {% for method in ["fill", "fill_override", "fill_override_intrinsics", "fill_override_intrinsics_bitwise"] %}
  test = start.dup

  x.report({{method}}) do
    guide.each { |value| test.{{method.id}} value }
  end

  x.report("{{method.id}}_only_true") do
    guide.size.times { test.{{method.id}} true }
  end

  x.report("{{method.id}}_only_false") do
    guide.size.times { test.{{method.id}} false }
  end
  {% end %}
end