Test JIT optimizer

output.log

>elixir test.exs
Operating System: Linux
CPU Information: AMD Ryzen 9 6900HS with Radeon Graphics
Number of Available Cores: 16
Available memory: 30.58 GB
Elixir 1.17.0
Erlang 26.2.3
JIT enabled: true

Benchmark suite executing with the following configuration:
warmup: 2 s
time: 5 s
memory time: 2 s
reduction time: 0 ns
parallel: 1
inputs: none specified
Estimated total run time: 18 s

Benchmarking dot_product_tco ...
Benchmarking dot_product_tco_guard ...
Calculating statistics...
Formatting results...

Name                            ips        average  deviation         median         99th %
dot_product_tco_guard       79.34 K       12.60 μs    ±87.28%       12.19 μs       19.47 μs
dot_product_tco             38.66 K       25.87 μs    ±59.17%       24.64 μs       51.80 μs

Comparison: 
dot_product_tco_guard       79.34 K
dot_product_tco             38.66 K - 2.05x slower +13.26 μs

Memory usage statistics:

Name                     Memory usage
dot_product_tco_guard           24 KB
dot_product_tco                  0 KB - 0.00x memory usage -24 KB

**All measurements for memory usage were the same**

test.exs

#!/usr/bin/env elixir
Mix.install([{:benchee, "~> 1.0"}])

defmodule Test do
  def benchmark() do
    vector1 = Enum.map(1..1536, fn _ -> :rand.uniform() end)
    vector2 = Enum.map(1..1536, fn _ -> :rand.uniform() end)

    Benchee.run(
      %{
        "dot_product_tco" => fn -> dot_product_tco(vector1, vector2) end,
        "dot_product_tco_guard" => fn -> dot_product_tco_guard(vector1, vector2) end
      },
      time: 5,
      memory_time: 2
    )
  end

  def dot_product_tco(vec1, vec2, acc \\ 0)
  def dot_product_tco([], [], acc), do: acc

  def dot_product_tco([x | vector1], [y | vector2], acc) do
    dot_product_tco(vector1, vector2, x * y + acc)
  end

  def dot_product_tco_guard(vec1, vec2, acc \\ 0)
  def dot_product_tco_guard([], [], acc), do: acc

  def dot_product_tco_guard([x | vector1], [y | vector2], acc) when is_float(x) and is_float(y) do
    dot_product_tco_guard(vector1, vector2, x * y + acc)
  end
end

Test.benchmark()