4 [![Build Status](https://travis-ci.org/Amanieu/parking_lot.svg?branch=master)](https://travis-ci.org/Amanieu/parking_lot) [![Build status](https://ci.appveyor.com/api/projects/status/wppcc32ttpud0a30/branch/master?svg=true)](https://ci.appveyor.com/project/Amanieu/parking-lot/branch/master) [![Crates.io](https://img.shields.io/crates/v/parking_lot.svg)](https://crates.io/crates/parking_lot)
6 [Documentation (synchronization primitives)](https://docs.rs/parking_lot/)
8 [Documentation (core parking lot API)](https://docs.rs/parking_lot_core/)
10 [Documentation (type-safe lock API)](https://docs.rs/lock_api/)
12 This library provides implementations of `Mutex`, `RwLock`, `Condvar` and
13 `Once` that are smaller, faster and more flexible than those in the Rust
14 standard library, as well as a `ReentrantMutex` type which supports recursive
15 locking. It also exposes a low-level API for creating your own efficient
16 synchronization primitives.
18 When tested on x86_64 Linux, `parking_lot::Mutex` was found to be 1.5x
19 faster than `std::sync::Mutex` when uncontended, and up to 5x faster when
20 contended from multiple threads. The numbers for `RwLock` vary depending on
21 the number of reader and writer threads, but are almost always faster than
22 the standard library `RwLock`, and even up to 50x faster in some cases.
26 The primitives provided by this library have several advantages over those
27 in the Rust standard library:
29 1. `Mutex` and `Once` only require 1 byte of storage space, while `Condvar`
30 and `RwLock` only require 1 word of storage space. On the other hand the
31 standard library primitives require a dynamically allocated `Box` to hold
32 OS-specific synchronization primitives. The small size of `Mutex` in
33 particular encourages the use of fine-grained locks to increase
35 2. Since they consist of just a single atomic variable, have constant
36 initializers and don't need destructors, these primitives can be used as
37 `static` global variables. The standard library primitives require
38 dynamic initialization and thus need to be lazily initialized with
40 3. Uncontended lock acquisition and release is done through fast inline
41 paths which only require a single atomic operation.
42 4. Microcontention (a contended lock with a short critical section) is
43 efficiently handled by spinning a few times while trying to acquire a
45 5. The locks are adaptive and will suspend a thread after a few failed spin
46 attempts. This makes the locks suitable for both long and short critical
48 6. `Condvar`, `RwLock` and `Once` work on Windows XP, unlike the standard
49 library versions of those types.
50 7. `RwLock` takes advantage of hardware lock elision on processors that
51 support it, which can lead to huge performance wins with many readers.
52 8. `RwLock` uses a task-fair locking policy, which avoids reader and writer
53 starvation, whereas the standard library version makes no guarantees.
54 9. `Condvar` is guaranteed not to produce spurious wakeups. A thread will
55 only be woken up if it timed out or it was woken up by a notification.
56 10. `Condvar::notify_all` will only wake up a single thread and requeue the
57 rest to wait on the associated `Mutex`. This avoids a thundering herd
58 problem where all threads try to acquire the lock at the same time.
59 11. `RwLock` supports atomically downgrading a write lock into a read lock.
60 12. `Mutex` and `RwLock` allow raw unlocking without a RAII guard object.
61 13. `Mutex<()>` and `RwLock<()>` allow raw locking without a RAII guard
63 14. `Mutex` and `RwLock` support [eventual fairness](https://trac.webkit.org/changeset/203350)
64 which allows them to be fair on average without sacrificing performance.
65 15. A `ReentrantMutex` type which supports recursive locking.
66 16. An *experimental* deadlock detector that works for `Mutex`,
67 `RwLock` and `ReentrantMutex`. This feature is disabled by default and
68 can be enabled via the `deadlock_detection` feature.
69 17. `RwLock` supports atomically upgrading an "upgradable" read lock into a
71 18. Optional support for [serde](https://docs.serde.rs/serde/). Enable via the
72 feature `serde`. **NOTE!** this support is for `Mutex`, `ReentrantMutex`,
73 and `RwLock` only; `Condvar` and `Once` are not currently supported.
77 To keep these primitives small, all thread queuing and suspending
78 functionality is offloaded to the *parking lot*. The idea behind this is
79 based on the Webkit [`WTF::ParkingLot`](https://webkit.org/blog/6161/locking-in-webkit/)
80 class, which essentially consists of a hash table mapping of lock addresses
81 to queues of parked (sleeping) threads. The Webkit parking lot was itself
82 inspired by Linux [futexes](http://man7.org/linux/man-pages/man2/futex.2.html),
83 but it is more powerful since it allows invoking callbacks while holding a queue
88 There are a few restrictions when using this library on stable Rust:
90 - You will have to use the `const_*` functions (e.g. `const_mutex(val)`) to
91 statically initialize the locking primitives. Using e.g. `Mutex::new(val)`
92 does not work on stable Rust yet.
93 - `RwLock` will not be able to take advantage of hardware lock elision for
94 readers, which improves performance when there are multiple readers.
96 To enable nightly-only functionality, you need to enable the `nightly` feature
101 Add this to your `Cargo.toml`:
108 To enable nightly-only features, add this to your `Cargo.toml` instead:
112 parking_lot = { version = "0.10", features = ["nightly"] }
115 The experimental deadlock detector can be enabled with the
116 `deadlock_detection` Cargo feature.
118 The core parking lot API is provided by the `parking_lot_core` crate. It is
119 separate from the synchronization primitives in the `parking_lot` crate so that
120 changes to the core API do not cause breaking changes for users of `parking_lot`.
122 ## Minimum Rust version
124 The current minimum required Rust version is 1.36. Any change to this is
125 considered a breaking change and will require a major version bump.
129 Licensed under either of
131 * Apache License, Version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or http://www.apache.org/licenses/LICENSE-2.0)
132 * MIT license ([LICENSE-MIT](LICENSE-MIT) or http://opensource.org/licenses/MIT)
138 Unless you explicitly state otherwise, any contribution intentionally submitted
139 for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any
140 additional terms or conditions.