The Cost of Concurrency Coordination with Jon Gjengset

Jon Gjengset explores the hardware-level costs of multi-threaded coordination, from CPU cache lines to Amdahl's law.

In this talk, Jon Gjengset explores the true cost of concurrency coordination – from Amdahl's law down to CPU cache lines – and what we can do about it. Modern hardware is moving to more cores rather than faster ones. For application developers, this means that further performance gains require _parallelism_; making progress on many tasks at the same time. But as anyone with experience writing multi-threaded code will tell you, this is easier said than done — those threads inevitably have to coordinate, and coordination is "expensive." But _why_ is it expensive? Sure, a mutual exclusion lock forces sequential execution, which limits overall speedup (per Amdahl's law). But is the lock itself expensive? Do reader-writer locks help when the coordination is mostly required for reads? In this talk, we'll dive deep into what makes concurrency coordination costly, and explore some pathways to mitigate that cost. By the end, you'll leave with a mental model that goes all the way down to the CPU cache line, and a newfound appreciation for a protocol from the 1980s. Chapters 00:05 Talk begins 36:19 Audience Q&A