mergerfs/libfuse/lib/pool.hpp

#pragma once

#include <tuple>
#include <atomic>
#include <vector>
#include <thread>
#include <memory>
#include <future>
#include <utility>
#include <stdexcept>
#include <functional>
#include <type_traits>
#include "queue.hpp"



class simple_thread_pool
{
public:
	explicit simple_thread_pool(std::size_t thread_count = std::thread::hardware_concurrency())
	{
		if(!thread_count)
			throw std::invalid_argument("bad thread count! must be non-zero!");

		auto worker = [this]()
		{
			while(true)
			{
				proc_t f;
				if(!m_queue.pop(f))
					break;
				f();
			}
		};

		m_threads.reserve(thread_count);
		while(thread_count--)
			m_threads.emplace_back(worker);
	}

	~simple_thread_pool()
	{
		m_queue.unblock();
		for(auto& thread : m_threads)
			thread.join();
	}

	template<typename F, typename... Args>
	void enqueue_work(F&& f, Args&&... args)
	{
		m_queue.push([p = std::forward<F>(f), t = std::make_tuple(std::forward<Args>(args)...)]() { std::apply(p, t); });
	}

	template<typename F, typename... Args>
	[[nodiscard]] auto enqueue_task(F&& f, Args&&... args) -> std::future<std::invoke_result_t<F, Args...>>
	{
		using task_return_type = std::invoke_result_t<F, Args...>;
		using task_type = std::packaged_task<task_return_type()>;

		auto task = std::make_shared<task_type>(std::bind(std::forward<F>(f), std::forward<Args>(args)...));
		auto result = task->get_future();

		m_queue.push([=]() { (*task)(); });

		return result;
	}

private:
	using proc_t = std::function<void(void)>;
	using queue_t = unbounded_queue<proc_t>;
	queue_t m_queue;

	using threads_t = std::vector<std::thread>;
	threads_t m_threads;
};



class thread_pool
{
public:
	explicit thread_pool(std::size_t thread_count = std::thread::hardware_concurrency())
	: m_queues(thread_count), m_count(thread_count)
	{
		if(!thread_count)
			throw std::invalid_argument("bad thread count! must be non-zero!");

		auto worker = [this](auto i)
		{
			while(true)
			{
				proc_t f;
				for(std::size_t n = 0; n < m_count * K; ++n)
					if(m_queues[(i + n) % m_count].try_pop(f))
						break;
				if(!f && !m_queues[i].pop(f))
					break;
				f();
			}
		};

		m_threads.reserve(thread_count);
		for(std::size_t i = 0; i < thread_count; ++i)
			m_threads.emplace_back(worker, i);
	}

	~thread_pool()
	{
		for(auto& queue : m_queues)
			queue.unblock();
		for(auto& thread : m_threads)
			thread.join();
	}

	template<typename F, typename... Args>
	void enqueue_work(F&& f, Args&&... args)
	{
		auto work = [p = std::forward<F>(f), t = std::make_tuple(std::forward<Args>(args)...)]() { std::apply(p, t); };
		auto i = m_index++;

		for(std::size_t n = 0; n < m_count * K; ++n)
			if(m_queues[(i + n) % m_count].try_push(work))
				return;

		m_queues[i % m_count].push(std::move(work));
	}

	template<typename F, typename... Args>
	[[nodiscard]] auto enqueue_task(F&& f, Args&&... args) -> std::future<std::invoke_result_t<F, Args...>>
	{
		using task_return_type = std::invoke_result_t<F, Args...>;
		using task_type = std::packaged_task<task_return_type()>;

		auto task = std::make_shared<task_type>(std::bind(std::forward<F>(f), std::forward<Args>(args)...));
		auto work = [=]() { (*task)(); };
		auto result = task->get_future();
		auto i = m_index++;

		for(auto n = 0; n < m_count * K; ++n)
			if(m_queues[(i + n) % m_count].try_push(work))
				return result;

		m_queues[i % m_count].push(std::move(work));

		return result;
	}

private:
	using proc_t = std::function<void(void)>;
	using queue_t = unbounded_queue<proc_t>;
	using queues_t = std::vector<queue_t>;
	queues_t m_queues;

	using threads_t = std::vector<std::thread>;
	threads_t m_threads;

	const std::size_t m_count;
	std::atomic_uint m_index = 0;

	inline static const unsigned int K = 2;
};