= Benchmarks

__Run on 11th Gen Intel(R) Core(TM) i7-1185G7 @ 3.00GHz__

== Posting to an executor

The benchmark is running the following code, with cobalt's task, `asio::awaitable` and `asio`'s
stackful coroutine (boost.context) based.

[source,cpp]
----
cobalt::task<void> atest()
{
  for (std::size_t i = 0u; i < n; i++)
    co_await asio::post(cobalt::use_op);
}
----

.results for 50M times in ms
[cols="1,1,1"]
|===
| |gcc 12 |clang 16

|cobalt     | 2472 | 2098
|awaitable | 2432 | 2253
|stackful  | 3655 | 3725
|===

== Running noop coroutine in parallel

This benchmark uses an `asio::experimental::channel` that has a size of zero,
to read & write in parallel to it. It uses <<gather, gather>> with cobalt
and an `awaitable_operator` in the `asio::awaitable`.

[source,cpp]
----
cobalt::task<void> atest()
{
  asio::experimental::channel<void(system::error_code)> chan{co_await cobalt::this_coro::executor, 0u};
  for (std::size_t i = 0u; i < n; i++)
    co_await cobalt::gather(
              chan.async_send(system::error_code{}, cobalt::use_task),
              chan.async_receive(cobalt::use_task));
}

asio::awaitable<void> awtest()
{
  asio::experimental::channel<void(system::error_code)> chan{co_await cobalt::this_coro::executor, 0u};
  using boost::asio::experimental::awaitable_operators::operator&&;
  for (std::size_t i = 0u; i < n; i++)
    co_await (
        chan.async_send(system::error_code{}, asio::use_awaitable)
        &&
        chan.async_receive(asio::use_awaitable));
}
----


.results for 3M times in ms
[cols="1,1,1"]
|===
| |gcc 12 |clang 16

|cobalt    | 1563 | 1468
|awaitable | 2800 | 2805
|===

== Immediate

This benchmark utilizes the immediate completion, by using a channel
with a size of 1, so that every operation is immediate.

[source,cpp]
----
cobalt::task<void> atest()
{
  asio::experimental::channel<void(system::error_code)> chan{co_await cobalt::this_coro::executor, 1u};
  for (std::size_t i = 0u; i < n; i++)
  {
    co_await chan.async_send(system::error_code{}, cobalt::use_op);
    co_await chan.async_receive(cobalt::use_op);
  }
}
----


.result for 10M times in ms
[cols="1,1,1"]
|===
| |gcc 12 |clang 16

|cobalt    | 1810 | 1864
|awaitable | 3109 | 4110
|stackful  | 3922 | 4705

|===

== Channels

In this benchmark asio::experimental::channel and cobalt::channel get compared.

This is similar to the parallel test, but uses the `cobalt::channel` instead.

.result of running the test 3M times in ms
[cols="1,1,1"]
|===
| |gcc |clang

|cobalt    | 500 | 350
|awaitable | 790 | 770
|stackful  | 867 | 907

|===

== Operation Allocations

This benchmark compares the different possible solutions for the associated allocator of asynchronous operations

.result of running the test 2M times in ms
[cols="1,1,1"]
|===
| |gcc |clang

|std::allocator    | 1136 | 1139
|cobalt::monotonic | 1149 | 1270
|pmr::monotonic    | 1164 | 1173
|cobalt::sbo       | 1021 | 1060

The latter method is used internally by cobalt.

|===