rtrb

rtrb (Real-Time Ring Buffer) is the leading wait-free SPSC queue for Rust, designed for hard real-time use cases like audio DSP and motor control. Push and pop are both wait-free with ~7 ns per operation on x86 and benchmarks reaching 520–530M ops/s on Apple M4 — placing it at the top of the published SPSC tier regardless of language.

Design

A bounded power-of-two ring-buffer with 128-byte padding between the producer and consumer indices, acquire/release atomics rather than seq_cst, and shadow variables (the producer caches the consumer's head, and vice versa) to convert per-operation cache-line transfers into per-batch transfers. Standard SPSC technique applied without compromise.

The API is split: constructing the queue returns a (Producer, Consumer) pair, and Send is implemented for each side but Sync is not. The split-handle design enforces the SPSC contract at the type level — there is no way to accidentally push from two threads.

Comparison

For MPMC use crossbeam ArrayQueue or kanal; for message-passing channels crossbeam-channel or kanal. For raw SPSC throughput, rtrb.

When to use it

• Audio and DSP pipelines (the original motivation)
• Telemetry / log buffers between exactly one producer and one consumer
• thread-per-core mailbox between adjacent shards
• Anywhere you can guarantee single-producer / single-consumer topology

Cross-language context

Rust's SPSC ring buffers (rtrb, ringbuffer-spsc, bounded-spsc-queue) produce essentially identical machine code to C++ equivalents (Rigtorp's SPSCQueue, folly's ProducerConsumerQueue, atomic_queue's SPSC mode) and achieve comparable latencies. Rust's advantage at this layer is safety, not speed: the Producer/Consumer split in rtrb statically prevents SPSC violations at compile time, eliminating a bug class that has plagued C++ and Java lock-free code for decades. The Java analogue (JCTools' SpscArrayQueue) uses Unsafe-based field access to match native performance, also at ~50 ns/op. See mechanical-sympathy for why all three converge.

See also

• spsc-queue — the optimization techniques rtrb applies
• ring-buffer — the underlying structure
• the-fastest-queue — full hierarchy
• inter-thread-communication — broader latency context
• mechanical-sympathy — why language barely matters here