Lock-free Queues in the Multiverse of Madness - Dave Rowland - C++Online 2026

Lock-free queues are the unsung heroes of real-time software systems, quietly enabling ultra-low latency and thread-safe communication across a dizzying variety of scenarios. But just like navigating a multiverse, the world of lock-free queues is filled with many different variants-each with their own quirks, trade-offs, and hardware demands. In this talk, we’ll embark on a journey through this multiverse, starting with the simplest dimension: the single-producer, single-consumer queue. We’ll uncover why lock-free designs often outshine traditional lock-based approaches in real-time environments, and explore various implementations along with their pros and cons. From there, we’ll warp into more complex realities featuring single-producer, multi-consumer and multi-producer, multi-consumer queues. Along the way, we’ll tackle mind-bending concepts like memory ordering, cache coherence, and platform-specific behaviour. By the end of this adventure, you’ll have the insight needed to confidently select and implement the perfect lock-free queue for your projects-no matter how mad the multiverse gets.