It’s not guaranteed. You can use the lock xadd instruction to achieve the same effect, or use C++ std::atomic, or use #pragma omp atomic, or any number of other concurrency solutions that have been written to save you the trouble of reinventing the wheel.
