Atomics

Until now, I've used two strategies to summate a std::vector. First, I did the whole math in one thread (Single Threaded: Summation of a vector); second multiple threads shared the same variable for the result (Multithreaded: Summation of a vector). In particular, the second strategy was extremely naive. In this post, I will apply my knowledge of both posts. My goal is that the thread will perform their summation as independently from each other as possible and therefore reduce the synchronization overhead. 

What is the fastest way to add the elements of a std::vector? This a question that I will pursue in the following posts. I use the single-threaded addition as the reference number. In further posts, I discuss atomics, locks, and thread-local data.

There are a lot of issues with the singleton pattern. I'm aware of that. But the singleton pattern is an ideal use case for a variable, which can only be initialized in a thread-safe way. From that point on, you can use it without synchronization. So in this post, I discuss different ways to initialize a singleton in a multithreading environment. You get the performance numbers and can reason about your use cases for the thread-safe initialization of a variable.

With the relaxed semantics, we have no synchronizations and ordering constraints on atomic operations.

But we can improve and further improve the acquire-release semantics of the last post. Why should x be atomic? There is no reason. That was my first but incorrect assumption. See why?

With the acquire-release semantics, we break the sequential consistency. In the acquire-release semantics, synchronization occurs between atomic operations on the same atomic and not between threads.

With atomic data types, you can tailor your program to your needs and optimize it. But now we are in the domain of multithreading experts.

CppMem is an interactive tool for exploring the behavior of small code snippets of the C++ memory model. No, it should have to be in the toolbox of each programmer who deals seriously with the memory model.

The relaxed semantics is the end of the scale. The relaxed semantic is the weakest of all memory models and guarantees that the operations on atomic variables are atomic.

Acquire and release fences guarantee similar synchronization and ordering constraints as atomics with acquire-release semantics. Similar because the differences are in the details.