Struct SharedRingMpscFifo 

pub struct SharedRingMpscFifo;

Factory for a single-ring MPSC pool that preserves global FIFO ordering across all producers. Producers contend on one Vyukov-MPMC producer_seq (CAS retries on push) while the single consumer skips the consumer-side CAS by using the try_pop_spsc fast path on the underlying SharedRing.

When this beats SharedRingMpsc (the composed variant)

SharedRingMpsc is faster on the producer side for any N because each producer owns its own ring (zero CAS contention). SharedRingMpscFifo wins on the consumer side because the consumer drains one ring instead of round-robining N. The crossover depends on:

• N (producer count): low N favours Fifo (light producer CAS contention + no consumer round-robin); high N favours Composed (independent producer rings, no shared CAS).
• ordering requirement: Fifo is the only choice when the caller needs global FIFO across all producers (e.g. a totally- ordered event log). Composed only preserves per-producer FIFO.

examples/mpmc_shootout.rs benches both side by side at common (N producers, 1 consumer) shapes; pick by measurement.

Implementations

impl SharedRingMpscFifo

pub fn create_anon_pool( n_producers: usize, capacity: usize, ) -> Result<(Vec<MpscFifoProducer>, MpscFifoConsumer), RingError>

Anonymous in-memory single-ring MPSC pool.

pub fn create_pool( path: impl AsRef<Path>, n_producers: usize, capacity: usize, ) -> Result<(Vec<MpscFifoProducer>, MpscFifoConsumer), RingError>

File-backed single-ring MPSC pool. One backing file (not one per producer), so cross-process layout is identical to a single SharedRing.

pub fn open_pool( path: impl AsRef<Path>, n_producers: usize, expected_capacity: usize, ) -> Result<(Vec<MpscFifoProducer>, MpscFifoConsumer), RingError>

Open an existing file-backed single-ring MPSC pool.