193 lines
6.2 KiB
C++
193 lines
6.2 KiB
C++
/*
|
|
Copyright (c) 2005-2022 Intel Corporation
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
*/
|
|
|
|
#ifndef __TBB_queuing_mutex_H
|
|
#define __TBB_queuing_mutex_H
|
|
|
|
#include "detail/_namespace_injection.h"
|
|
#include "detail/_assert.h"
|
|
#include "detail/_utils.h"
|
|
#include "detail/_mutex_common.h"
|
|
|
|
#include "profiling.h"
|
|
|
|
#include <atomic>
|
|
|
|
namespace tbb {
|
|
namespace detail {
|
|
namespace d1 {
|
|
|
|
//! Queuing mutex with local-only spinning.
|
|
/** @ingroup synchronization */
|
|
class queuing_mutex {
|
|
public:
|
|
//! Construct unacquired mutex.
|
|
queuing_mutex() noexcept {
|
|
create_itt_sync(this, "tbb::queuing_mutex", "");
|
|
};
|
|
|
|
queuing_mutex(const queuing_mutex&) = delete;
|
|
queuing_mutex& operator=(const queuing_mutex&) = delete;
|
|
|
|
//! The scoped locking pattern
|
|
/** It helps to avoid the common problem of forgetting to release lock.
|
|
It also nicely provides the "node" for queuing locks. */
|
|
class scoped_lock {
|
|
//! Reset fields to mean "no lock held".
|
|
void reset() {
|
|
m_mutex = nullptr;
|
|
}
|
|
|
|
public:
|
|
//! Construct lock that has not acquired a mutex.
|
|
/** Equivalent to zero-initialization of *this. */
|
|
scoped_lock() = default;
|
|
|
|
//! Acquire lock on given mutex.
|
|
scoped_lock(queuing_mutex& m) {
|
|
acquire(m);
|
|
}
|
|
|
|
//! Release lock (if lock is held).
|
|
~scoped_lock() {
|
|
if (m_mutex) release();
|
|
}
|
|
|
|
//! No Copy
|
|
scoped_lock( const scoped_lock& ) = delete;
|
|
scoped_lock& operator=( const scoped_lock& ) = delete;
|
|
|
|
//! Acquire lock on given mutex.
|
|
void acquire( queuing_mutex& m ) {
|
|
__TBB_ASSERT(!m_mutex, "scoped_lock is already holding a mutex");
|
|
|
|
// Must set all fields before the exchange, because once the
|
|
// exchange executes, *this becomes accessible to other threads.
|
|
m_mutex = &m;
|
|
m_next.store(nullptr, std::memory_order_relaxed);
|
|
m_going.store(0U, std::memory_order_relaxed);
|
|
|
|
// x86 compare exchange operation always has a strong fence
|
|
// "sending" the fields initialized above to other processors.
|
|
scoped_lock* pred = m.q_tail.exchange(this);
|
|
if (pred) {
|
|
call_itt_notify(prepare, &m);
|
|
__TBB_ASSERT(pred->m_next.load(std::memory_order_relaxed) == nullptr, "the predecessor has another successor!");
|
|
|
|
pred->m_next.store(this, std::memory_order_release);
|
|
spin_wait_while_eq(m_going, 0U);
|
|
}
|
|
call_itt_notify(acquired, &m);
|
|
|
|
}
|
|
|
|
//! Acquire lock on given mutex if free (i.e. non-blocking)
|
|
bool try_acquire( queuing_mutex& m ) {
|
|
__TBB_ASSERT(!m_mutex, "scoped_lock is already holding a mutex");
|
|
|
|
// Must set all fields before the compare_exchange_strong, because once the
|
|
// compare_exchange_strong executes, *this becomes accessible to other threads.
|
|
m_next.store(nullptr, std::memory_order_relaxed);
|
|
m_going.store(0U, std::memory_order_relaxed);
|
|
|
|
scoped_lock* expected = nullptr;
|
|
// The compare_exchange_strong must have release semantics, because we are
|
|
// "sending" the fields initialized above to other processors.
|
|
// x86 compare exchange operation always has a strong fence
|
|
if (!m.q_tail.compare_exchange_strong(expected, this, std::memory_order_acq_rel))
|
|
return false;
|
|
|
|
m_mutex = &m;
|
|
|
|
call_itt_notify(acquired, &m);
|
|
return true;
|
|
}
|
|
|
|
//! Release lock.
|
|
void release()
|
|
{
|
|
__TBB_ASSERT(this->m_mutex, "no lock acquired");
|
|
|
|
call_itt_notify(releasing, this->m_mutex);
|
|
|
|
if (m_next.load(std::memory_order_relaxed) == nullptr) {
|
|
scoped_lock* expected = this;
|
|
if (m_mutex->q_tail.compare_exchange_strong(expected, nullptr)) {
|
|
// this was the only item in the queue, and the queue is now empty.
|
|
reset();
|
|
return;
|
|
}
|
|
// Someone in the queue
|
|
spin_wait_while_eq(m_next, nullptr);
|
|
}
|
|
m_next.load(std::memory_order_acquire)->m_going.store(1U, std::memory_order_release);
|
|
|
|
reset();
|
|
}
|
|
|
|
private:
|
|
//! The pointer to the mutex owned, or nullptr if not holding a mutex.
|
|
queuing_mutex* m_mutex{nullptr};
|
|
|
|
//! The pointer to the next competitor for a mutex
|
|
std::atomic<scoped_lock*> m_next{nullptr};
|
|
|
|
//! The local spin-wait variable
|
|
/** Inverted (0 - blocked, 1 - acquired the mutex) for the sake of
|
|
zero-initialization. Defining it as an entire word instead of
|
|
a byte seems to help performance slightly. */
|
|
std::atomic<uintptr_t> m_going{0U};
|
|
};
|
|
|
|
// Mutex traits
|
|
static constexpr bool is_rw_mutex = false;
|
|
static constexpr bool is_recursive_mutex = false;
|
|
static constexpr bool is_fair_mutex = true;
|
|
|
|
private:
|
|
//! The last competitor requesting the lock
|
|
std::atomic<scoped_lock*> q_tail{nullptr};
|
|
|
|
};
|
|
|
|
#if TBB_USE_PROFILING_TOOLS
|
|
inline void set_name(queuing_mutex& obj, const char* name) {
|
|
itt_set_sync_name(&obj, name);
|
|
}
|
|
#if (_WIN32||_WIN64)
|
|
inline void set_name(queuing_mutex& obj, const wchar_t* name) {
|
|
itt_set_sync_name(&obj, name);
|
|
}
|
|
#endif //WIN
|
|
#else
|
|
inline void set_name(queuing_mutex&, const char*) {}
|
|
#if (_WIN32||_WIN64)
|
|
inline void set_name(queuing_mutex&, const wchar_t*) {}
|
|
#endif //WIN
|
|
#endif
|
|
} // namespace d1
|
|
} // namespace detail
|
|
|
|
inline namespace v1 {
|
|
using detail::d1::queuing_mutex;
|
|
} // namespace v1
|
|
namespace profiling {
|
|
using detail::d1::set_name;
|
|
}
|
|
} // namespace tbb
|
|
|
|
#endif /* __TBB_queuing_mutex_H */
|