benchmarking map serializoation

benchit.ex

defmodule Benchit do

  def run do
    ml = for i<- 1..1_000_000, do: {i,[]}, into: %{}
    mt = for i<- 1..1_000_000, do: {i,true}, into: %{}
    bl = :erlang.term_to_binary ml
    bt = :erlang.term_to_binary mt

    Benchee.run %{
      "serialize empty list" => fn -> :erlang.term_to_binary ml end,
      "serialize true" => fn -> :erlang.term_to_binary mt end,
      "deserialize empty list" => fn -> :erlang.binary_to_term bl end,
      "deserialize true" => fn -> :erlang.binary_to_term bt end,
    }, time: 10, parallel: 4
  end
end

out.txt

$ MIX_ENV=prod mix run -e 'Benchit.run'
Erlang/OTP 19 [erts-8.2.1] [source] [64-bit] [smp:28:28] [async-threads:10] [hipe] [kernel-poll:false]
Elixir 1.4.2
Benchmark suite executing with the following configuration:
warmup: 2.0s
time: 10.0s
parallel: 4
inputs: none specified
Estimated total run time: 48.0s

Benchmarking deserialize empty list...
Benchmarking deserialize true...
Benchmarking serialize empty list...
Benchmarking serialize true...

Name                             ips        average  deviation         median
serialize empty list           22.53       44.38 ms     ±3.77%       44.15 ms
serialize true                 18.35       54.51 ms     ±3.05%       54.46 ms
deserialize empty list          6.07      164.77 ms     ±7.63%      158.50 ms
deserialize true                4.14      241.48 ms     ±4.71%      235.95 ms

Comparison: 
serialize empty list           22.53
serialize true                 18.35 - 1.23x slower
deserialize empty list          6.07 - 3.71x slower
deserialize true                4.14 - 5.44x slower

tested on:
Intel(R) Xeon(R) CPU E5-2695 v3 @ 2.30GHz