Hi community! Wanted to share a project of mine that is now enabling parallelism across many of my open-source projects: ForkUnion. The absolute majority of "thread pool" projects on GitHub are `std::deque<std::shared_ptr<task_t>>` task queues guarded by a `std::mutex`. I'm not sure which part of that equation I'm more allergic to, but I knew I had to find an alternative. It probably exists somewhere, but I ended up implementing everything from scratch last year. This thread pool is designed around hardware properties. It avoids dynamic memory allocations (obviously), mutexes (less obviously), and even Compare-And-Swap atomics (what?!). It's designed exclusively around Read-Modify-Write atomics and basic `for_n` and `for_n_dynamic` verbs. A very minimalistic vocabulary with a lot of machinery under the hood. Namely, it can leverage WFET on Arm and Zawrs on RISC-V, it knows how to demote cache lines on x86, and it tracks the NUMA topology and measures compute-domain to memory-domain affinity itself, instead of relying on the OS... and a lot more! It is available for C, C++, Rust, and Zig. It passes CI on Linux, macOS, Windows, iOS, and Android, despite every one of those OSes having quite different ways of exposing and controlling threads and memory. It also comes with runtime dynamic dispatch (like all of my major projects), so you can compile once, ship your application everywhere, and it will automatically leverage the newest capabilities on that platform. Thought it was a good time to share it here. Let me know if you notice any bugs or have features in mind ;) |