Learning C++ Memory Model from a Distributed System's Perspective (2021)

The article discusses understanding the C++ memory model from a distributed system's perspective. It highlights how C++ provides abstractions for writing software that runs on distributed systems through std::memory_order. The memory model in C++ includes various types of happens-before relationships and orderings like release-consume, release-acquire, and sequentially-consistent ordering. The happens-before relationship establishes a partial order consistent with the evaluation order. The article explains memory_order_relaxed, synchronization operations, and the significance of release-acquire ordering in ensuring visibility of memory writes across threads. It compares using release and acquire operations to achieve eventual consistency similar to reading from a primary in a distributed key-value store. Additionally, it mentions memory_order_seq_cst as the strongest and default memory order, providing total ordering and global linearizability across multiple std::atomics. Overall, the article delves into how C++ memory model concepts align with principles observed in distributed systems, offering insights into memory ordering and synchronization in multi-threaded environments.