rpcs3/Utilities/cond.h

#pragma once

#include "types.h"
#include "Atomic.h"
#include <shared_mutex>
#include "asm.h"

// Lightweight condition variable
class cond_variable
{
	// Internal waiter counter
	atomic_t<u32> m_value{0};

	friend class notifier;

protected:
	// Internal waiting function
	bool imp_wait(u32 _old, u64 _timeout) noexcept;

	// Try to notify up to _count threads
	void imp_wake(u32 _count) noexcept;

public:
	constexpr cond_variable() = default;

	// Intrusive wait algorithm for lockable objects
	template <typename T>
	bool wait(T& object, u64 usec_timeout = -1)
	{
		const u32 _old = m_value.fetch_add(1); // Increment waiter counter
		object.unlock();
		const bool res = imp_wait(_old, usec_timeout);
		object.lock();
		return res;
	}

	// Unlock all specified objects but don't lock them again
	template <typename... Locks>
	bool wait_unlock(u64 usec_timeout, Locks&&... locks)
	{
		const u32 _old = m_value.fetch_add(1); // Increment waiter counter
		(..., std::forward<Locks>(locks).unlock());
		return imp_wait(_old, usec_timeout);
	}

	// Wake one thread
	void notify_one() noexcept
	{
		if (m_value)
		{
			imp_wake(1);
		}
	}

	// Wake all threads
	void notify_all() noexcept
	{
		if (m_value)
		{
			imp_wake(65535);
		}
	}

	static constexpr u64 max_timeout = u64{UINT32_MAX} / 1000 * 1000000;
};

// Pair of a fake shared mutex (only limited shared locking) and a condition variable. Obsolete.
class notifier
{
	atomic_t<u32> m_counter{0};
	cond_variable m_cond;

	bool imp_try_lock(u32 count);

	void imp_unlock(u32 count);

	u32 imp_notify(u32 count);

public:
	constexpr notifier() = default;

	bool try_lock()
	{
		return imp_try_lock(max_readers);
	}

	void unlock()
	{
		imp_unlock(max_readers);
	}

	bool try_lock_shared()
	{
		return imp_try_lock(1);
	}

	void unlock_shared()
	{
		imp_unlock(1);
	}

	bool wait(u64 usec_timeout = -1);

	void notify_all()
	{
		if (m_counter)
		{
			imp_notify(-1);
		}

		// Notify after imaginary "exclusive" lock+unlock
		m_cond.notify_all();
	}

	void notify_one()
	{
		// TODO
		if (m_counter)
		{
			if (imp_notify(1))
			{
				return;
			}
		}

		m_cond.notify_one();
	}

	static constexpr u32 max_readers = 0x7f;
};

// Condition variable fused with a pseudo-mutex which is never supposed to be locked concurrently.
class unique_cond
{
	enum : u32
	{
		c_wait = 1,
		c_lock = 2,
		c_sig  = 3,
	};

	atomic_t<u32> m_value{0};

	bool imp_wait(u64 _timeout) noexcept;
	void imp_notify() noexcept;

public:
	constexpr unique_cond() = default;

	void lock() noexcept
	{
		// Shouldn't be locked by more than one thread concurrently
		while (UNLIKELY(!m_value.compare_and_swap_test(0, c_lock)))
			;
	}

	void unlock() noexcept
	{
		m_value = 0;
	}

	bool wait(std::unique_lock<unique_cond>& lock, u64 usec_timeout = -1) noexcept
	{
		AUDIT(lock.owns_lock());
		AUDIT(lock.mutex() == this);
		return imp_wait(usec_timeout);
	}

	void notify() noexcept
	{
		// Early exit if notification is not required
		if (LIKELY(!m_value))
			return;

		imp_notify();
	}
};

// Packed version of cond_one, supports up to 16 readers.
class cond_x16
{
	// For information, shouldn't modify
	enum : u32
	{
		c_wait = 1,
		c_lock = 1 << 16,
		c_sig  = 1 << 16 | 1,
	};

	// Split in 16-bit parts for convenient bit combining
	atomic_t<u32> m_cvx16{0};

	// Effectively unused, only counts readers
	atomic_t<u32> m_total{0};

	class lock_x16
	{
		cond_x16* m_this;
		u32 m_slot;

		friend class cond_x16;

	public:
		lock_x16(cond_x16* _this) noexcept
			: m_this(_this)
		{
			// Spin if the number of readers exceeds 16
			while (UNLIKELY(m_this->m_total++ >= 16))
				m_this->m_total--;

			// Lock and remember obtained slot index
			m_slot = m_this->m_cvx16.atomic_op([](u32& cvx16)
			{
				// Combine used bits and invert to find least significant bit unused
				const u32 slot = utils::cnttz32(~((cvx16 & 0xffff) | (cvx16 >> 16)), true);

				// Set lock bit
				cvx16 |= c_lock << slot;

				AUDIT(slot < 16);
				return slot;
			});
		}

		lock_x16(const lock_x16&) = delete;

		lock_x16& operator=(const lock_x16&) = delete;

		~lock_x16()
		{
			// Clear the slot
			m_this->m_cvx16 &= ~((c_wait | c_lock) << m_slot);
			m_this->m_total -= 1;
		}

		bool wait(u64 usec_timeout = -1) const noexcept
		{
			return m_this->wait(*this, usec_timeout);
		}
	};

	bool imp_wait(u32 slot, u64 _timeout) noexcept;
	void imp_notify() noexcept;

public:
	constexpr cond_x16() = default;

	lock_x16 lock_one() noexcept
	{
		return lock_x16(this);
	}

	bool wait(lock_x16 const& lock, u64 usec_timeout = -1) noexcept
	{
		AUDIT(lock.m_this == this);
		return imp_wait(lock.m_slot, usec_timeout);
	}

	void notify_all() noexcept
	{
		if (LIKELY(!m_cvx16))
			return;

		imp_notify();
	}
};