Concurrent collections and synchronization primitives for writing fast multithreaded and asynchronous code. This library is experimental and provided without warranty, so use it carefully.
- lockless semaphores
- priority semaphore
- lockless object and array pools
- memory compact single writer multiple readers hash map
- memory compact thread safe (but locked) hash map
- write optimized lock-free counter
QueueSemaphore: usesConcurrentQueueto order waitersStackSemaphore: usesConcurrentStackto order waitersSimpleSegmentSemaphore: uses segment queue to order waiters (similar to the Semaphore from kotlinx.coroutines)SegmentSemaphore: uses segment queue with support of early removal of cancelled waitersPrioritySemaphore: lock-based semaphore (like SemaphoreSlim), but releases waiters in specified priority
LocklessArrayPool<T>.Shared: similar toArrayPool<T>.Sharedbut uses bounded concurrent queue instead of lock protected arrays to store pooled items.IObjectPool<T>,ObjectPool<T>
SingleWriterDictionary<TKey, TValue, TComparer>: memory-compact hash map that supports single writer and multiple readers simultaneously. Writes are lock-free (except for resizes, which may trigger GC), but reads may be retried when a parallel update occurs.StripedDictionary<TKey, TValue, TComparer>: memory-compact hash map, designed to avoidConcurrentDictionaryGC overhead. It is not general purposeConcurrentDictionaryreplacement, and generally is slower, especially for reads.DefaultComparer<TKey>: default comparer for ConcurrencyToolkit hash mapsComparerWrapper<TKey>: struct wrapper over the reference type comparer
ThreadSafeCounter64: scalable atomic counter that distributes writes across per-core variables
InterlockedFloatingPoint: methods for atomic floating point arithmeticsLowLatencySpinWait:SpinWaitimplementation that achieves low latency (~0.5ms) on sleeps instead of standard 15ms
