Lock Free Queues
Public domain implementation of four different lock free queues:
- SPSC lock free dynamic queue which requires a memory allocation with each insert.
- MPSC lock free dynamic queue which requires a memory allocation with each insert.
- SPSC wait free bound queue/ring buffer which which uses a fixed size pre-allocated buffer.
- MPMC lock free bound queue/ring buffer which which uses a fixed size pre-allocated buffer.
Performance
These number are on my Early 2013 Retina Macbook Pro (2.7 GHz i7). The latency variability of memory allocation can have a signifigant impact on the dynamic queues, sometimes halving throughput.
| Queue | Throughput |
|---|---|
| SPSC dynamic | 8 million |
| MPSC dynamic | 8 million |
| SPSC fixed | 200 million |
| MPMC fixed | 54 million |