← Quit0 / 0 correct
RustHard1 of 266

A lock-free counter uses fetch_add on an AtomicU64 from 64 threads, and it's slower than expected. Why, and what helps?