Optimizations (#1680)

* Optimizations

1) Some headers simplified for better compilation time
2) Some templates simplified for smaller executable size
3) Eliminate std::future to fix compilation for mingw64
4) PKG installation can be cancelled now
5) cellGame fixes
6) XAudio2 fix for mingw64
7) PPUInterpreter bug fixed (Clang)

* any_pod<> implemented

Aliases: any16, any32, any64
rsx::make_command fixed

Utilities/SharedMutex.cpp

@@ -0,0 +1,157 @@
+#include "SharedMutex.h"
+
+#include <mutex>
+#include <condition_variable>
+
+struct shared_mutex::internal
+{
+	std::mutex mutex;
+
+	std::size_t rq_size{}; // Reader queue size (threads waiting on m_rcv)
+	std::size_t wq_size{}; // Writer queue size (threads waiting on m_wcv and m_ocv)
+
+	std::condition_variable rcv; // Reader queue
+	std::condition_variable wcv; // Writer queue
+	std::condition_variable ocv; // For current exclusive owner
+};
+
+shared_mutex::~shared_mutex()
+{
+	delete m_data;
+}
+
+void shared_mutex::initialize_once()
+{
+	if (!m_data)
+	{
+		auto ptr = new shared_mutex::internal;
+
+		if (!m_data.compare_and_swap_test(nullptr, ptr))
+		{
+			delete ptr;
+		}
+	}
+}
+
+void shared_mutex::lock_shared_hard()
+{
+	initialize_once();
+
+	std::unique_lock<std::mutex> lock(m_data->mutex);
+
+	// Validate
+	if ((m_ctrl & SM_INVALID_BIT) != 0) throw std::runtime_error("shared_mutex::lock_shared(): Invalid bit");
+	if ((m_ctrl & SM_READER_MASK) == 0) throw std::runtime_error("shared_mutex::lock_shared(): No readers");
+
+	// Notify non-zero reader queue size
+	m_ctrl |= SM_WAITERS_BIT, m_data->rq_size++;
+
+	// Fix excess reader count
+	if ((--m_ctrl & SM_READER_MASK) == 0 && m_data->wq_size)
+	{
+		// Notify exclusive owner
+		m_data->ocv.notify_one();
+	}
+
+	// Obtain the reader lock
+	while (true)
+	{
+		const auto ctrl = m_ctrl.load();
+
+		// Check writers and reader limit
+		if (m_data->wq_size || (ctrl & ~SM_WAITERS_BIT) >= SM_READER_MAX)
+		{
+			m_data->rcv.wait(lock);
+			continue;
+		}
+
+		if (m_ctrl.compare_and_swap_test(ctrl, ctrl + 1))
+		{
+			break;
+		}
+	}
+
+	if (!--m_data->rq_size && !m_data->wq_size)
+	{
+		m_ctrl &= ~SM_WAITERS_BIT;
+	}
+}
+
+void shared_mutex::unlock_shared_notify()
+{
+	initialize_once();
+
+	// Mutex is locked for reliable notification because m_ctrl has been changed outside
+	std::lock_guard<std::mutex> lock(m_data->mutex);
+
+	if ((m_ctrl & SM_READER_MASK) == 0 && m_data->wq_size)
+	{
+		// Notify exclusive owner
+		m_data->ocv.notify_one();
+	}
+	else if (m_data->rq_size)
+	{
+		// Notify other readers
+		m_data->rcv.notify_one();
+	}
+}
+
+void shared_mutex::lock_hard()
+{
+	initialize_once();
+
+	std::unique_lock<std::mutex> lock(m_data->mutex);
+
+	// Validate
+	if ((m_ctrl & SM_INVALID_BIT) != 0) throw std::runtime_error("shared_mutex::lock(): Invalid bit");
+
+	// Notify non-zero writer queue size
+	m_ctrl |= SM_WAITERS_BIT, m_data->wq_size++;
+
+	// Obtain the writer lock
+	while (true)
+	{
+		const auto ctrl = m_ctrl.load();
+
+		if (ctrl & SM_WRITER_LOCK)
+		{
+			m_data->wcv.wait(lock);
+			continue;
+		}
+
+		if (m_ctrl.compare_and_swap_test(ctrl, ctrl | SM_WRITER_LOCK))
+		{
+			break;
+		}
+	}
+
+	// Wait for remaining readers
+	while ((m_ctrl & SM_READER_MASK) != 0)
+	{
+		m_data->ocv.wait(lock);
+	}
+
+	if (!--m_data->wq_size && !m_data->rq_size)
+	{
+		m_ctrl &= ~SM_WAITERS_BIT;
+	}
+}
+
+void shared_mutex::unlock_notify()
+{
+	initialize_once();
+
+	// Mutex is locked for reliable notification because m_ctrl has been changed outside
+	std::lock_guard<std::mutex> lock(m_data->mutex);
+
+	if (m_data->wq_size)
+	{
+		// Notify next exclusive owner
+		m_data->wcv.notify_one();
+	}
+	else if (m_data->rq_size)
+	{
+		// Notify all readers
+		m_data->rcv.notify_all();
+	}
+}