|0.4.1||Jul 25, 2018|
|0.3.2||Jan 5, 2018|
|0.3.0||Aug 1, 2017|
|0.2.10||Aug 11, 2016|
|0.1.6||Nov 2, 2015|
#14 in Concurrency
82,581 downloads per month
Crossbeam supports concurrent programming, especially focusing on memory management, synchronization, and non-blocking data structures.
Crossbeam consists of several submodules:
AtomicConsumeprovides C/C++11-style "consume" atomic operations (re-exported from
ArcCellprovides atomic storage and retrieval of
threadfor utilities, re-exported from
crossbeam-utils. The "scoped" thread API in
threadmakes it possible to spawn threads that share stack data with their parents. The
utils::CachePaddedstruct inserts padding to align data with the size of a cacheline. This crate also seeks to expand the standard library's few synchronization primitives (locks, barriers, etc) to include advanced/niche primitives, as well as userspace alternatives.
epochfor memory management, re-exported from
crossbeam-epoch. Because non-blocking data structures avoid global synchronization, it is not easy to tell when internal data can be safely freed. The crate provides generic, easy to use, and high-performance APIs for managing memory in these cases. We plan to support other memory management schemes, e.g. hazard pointers (HP) and quiescent state-based reclamation (QSBR).
Concurrent data structures which are non-blocking and much superior to wrapping sequential ones with a
Mutex. Crossbeam currently provides channels (re-exported from
crossbeam-channel), deques (re-exported from
crossbeam-deque), queues, and stacks. Ultimately the goal is to also include bags, sets and maps.
To use Crossbeam, add this to your
[dependencies] crossbeam = "0.4"
For examples of what Crossbeam is capable of, see the documentation.