If you're being pedantic, atomic instructions are a misnomer. But "Lock free" is a property of parallel algorithms where if there's N > 1 processors evaluating

If you're being pedantic, atomic instructions are a misnomer. But "Lock free" is a property of parallel algorithms where if there's N > 1 processors evaluating an algorithm, there is no state where the algorithm cannot make progress ("locked" meaning "all N processors are waiting") Whether you want to debate if a physical system can realize a lock-free algorithm is a different debate. Hell, half the "well known" lock-free algorithms assume memory allocation/reclamation isn't blocking.