In this post, our tour through the c++ memory model goes one step deeper. Until now, the posts were only about the atomicity of the atomic data types but now we deal with the synchronisation and ordering constraints of the operations.
You can not configure the atomicity of an atomic data type, but you can adjust very accurately the synchronisation and ordering constraints of atomic operations. Leverage, which is unique to C++. That's not possible in the C#''s or Java's memory model.
The six variants of the C++ memory model
C++ has six variants of the memory model. The default for atomic operations is std::memory_order_seq_cst. But you can explicitly specify one of the other five. But what has C++11 to offer?
It helps a lot to answer two questions, to get a system into the six memory models.
- For which type of atomic operations should you use the memory model?
- Which synchronisation and ordering constraints are defined by the memory model?
The rest of this post is about answering these questions. So what are the types of atomic operations?
Modernes C++ Mentoring
Get the invitation to the one-hour presentation of my mentoring program "Fundamentals for C++ Professionals" including Q&A
- First: 2022-10-03; 9 pm (CEST)
- Second: 2022-10-10; 9 am (CEST)
Do you want the invitation to the Zoom meeting?
Types of atomic operations
The memory model deals with reading and/or writing atomic operations.
- read operation: memory_order_acquire and memory_order_consume
- write operation: memory_order_release
- read-modify-write operation: memory_order_acq_rel and memory_order_seq_cst
memory_order_relaxed defines no synchronisation and ordering constraints. So it will not fit in this taxonomy.
The table orders the atomic operations based on their reading and/or writing characteristics.
In case you use an atomic operation atomVar.load(5) with a memory model, that is designed for a write or read-modify-write operation, the write part has no effect. So an atomVar.load(5,std, std::memory_order_acq_rel) is equivalent to an atomVar.load(5,std::memory_order_acquire), an atomVar.load(5, std::memory_order_release) is equivalent to an atomVar.load(5, std::memory_order_relaxed).
The different synchronisation and ordering constraints
There are three different types of synchronization and ordering constraints in C++11:
- Sequential consistency: memory_order_seq_cst
- Acquire-release: memory_order_consume, memory_order_acquire, memory_order_release and memory_order_acq_rel
- Relaxed: memory_order_relaxed
While the sequential consistency establishes a global order between threads, the acquire-release semantic establishes an ordering between reading and write operations on the same atomic variable on different threads. The relaxed semantic only guarantees that operations on the same atomic data type in the same thread can not be reordered. That guarantee is called modification order consistency. But other threads can see this operation in a different order.
Admit, that was a short post. But I will stick to my idea, to talk only on one topic in one post.
Especially the levels of the different memory models, their effects on atomic and non-atomic operations, make the C++ memory model a thrilling but also challenging topic. In the next post, I will talk about the synchronization and ordering constraints of the sequential consistency, the acquire-release semantic, and the relaxed semantic. I will do it in theory and practice. The next post with be about the application of sequential consistency. The base is laid out in the post sequential consistency.
Thanks a lot to my Patreon Supporters: Matt Braun, Roman Postanciuc, Tobias Zindl, Marko, G Prvulovic, Reinhold Dröge, Abernitzke, Frank Grimm, Sakib, Broeserl, António Pina, Sergey Agafyin, Андрей Бурмистров, Jake, GS, Lawton Shoemake, Animus24, Jozo Leko, John Breland, Louis St-Amour, Venkat Nandam, Jose Francisco, Douglas Tinkham, Kuchlong Kuchlong, Robert Blanch, Truels Wissneth, Kris Kafka, Mario Luoni, Neil Wang, Friedrich Huber, lennonli, Pramod Tikare Muralidhara, Peter Ware, Daniel Hufschläger, Alessandro Pezzato, Evangelos Denaxas, Bob Perry, Satish Vangipuram, Andi Ireland, Richard Ohnemus, Michael Dunsky, Leo Goodstadt, John Wiederhirn, Yacob Cohen-Arazi, Florian Tischler, Robin Furness, Michael Young, Holger Detering, Bernd Mühlhaus, Matthieu Bolt, Stephen Kelley, Kyle Dean, Tusar Palauri, Dmitry Farberov, Juan Dent, George Liao, Daniel Ceperley, Jon T Hess, Stephen Totten, and Wolfgang Fütterer.
Thanks in particular to Jon Hess, Lakshman, Christian Wittenhorst, Sherhy Pyton, Dendi Suhubdy, Sudhakar Belagurusamy, Richard Sargeant, Rusty Fleming, Ralf Abramowitsch, John Nebel, Mipko, and Alicja Kaminska.
My special thanks to Embarcadero
My special thanks to PVS-Studio
I'm happy to give online seminars or face-to-face seminars worldwide. Please call me if you have any questions.
Standard Seminars (English/German)
Here is a compilation of my standard seminars. These seminars are only meant to give you a first orientation.