Concurrent Data Structures
High-performance C++20 concurrency library: a lock-free wait-free queue, a fine-grained concurrent hash map, and a work-stealing thread pool - plus the Eraser lockset algorithm for dynamic data race detection.
Sole engineer
- operation latency
- < 1 µs
- queue ops
- wait-free
- GoogleTest cases
- 59
- runtime deps
- 0
The challenge
Shared-state concurrency usually buys safety with locks and contention; this library builds the structures that scale across cores instead - lock-free and wait-free where that wins, fine-grained locking where it does not.
What I built
Three concurrent structures ship as header-only C++20 with zero runtime dependencies, building clean on GCC 10+, Clang 12+ and MSVC 2019+, covered by 59 GoogleTest cases and an ImGui monitor that graphs throughput, latency and queue depth while the benchmarks run. Wait-free queue operations and a work-stealing thread pool hold latency under a microsecond. Alongside them sits the Eraser lockset algorithm, implemented in C++ from the literature, to catch data races in multithreaded code dynamically instead of by inspection.
Have something in mind?
I'm open to backend and platform roles, contract or permanent. Based in the EU, available on-site in Belgium and the Netherlands or remotely worldwide.