performance and scalability 3.So Procrastinate away! Use Structured Deferral 4.Shared_ptr vs atomic_shared_ptr vs hazard pointers vs Read Copy Update (RCU) 5.Hazard Pointers 6.Read Copy Update 7.A Concurrency These are cell, hazard ptr and RCU. These extend the existing shared_ptr and the proposed atomic_shared_ptr which all have safe reclamation facilities. As such we also propose moving shared_ptr and atomic> These are cell, hazard ptr and RCU. These extend the existing shared_ptr and the proposed atomic_shared_ptr which all have safe reclamation facilities. As such we also propose moving shared_ptr and atomic>

Solution: ▪ C++11: Atomic operations for std::shared_ptr ▪ C++20: Atomic shared pointers ▪ std::atomic_shared_ptr ▪ std::atomic_weak_ptr Concurrency 23Atomic Smart Pointers Atomic smart pointers are part part of the C++20 standard. ▪ Partial specialization of std::atomic ▪ std::atomic_shared_ptr std::atomic> ▪ std::atomic_weak_ptr std::atomic> 24Smart Pointer ▪ Further

complete implementation of the C++ Concurrency TS, including a lock-free implementation of atomic_shared_ptr and RCU. http://stdthread.co.uk

Conclusion & Future work • My implementation is here: https://github.com/DanielLiamAnderson/atomic_shared_ptr • Still a proof-of-concept / prototype state • Need to implement atomic • Need

template class concurrent_stack { struct Node { T t; shared_ptr next; }; atomic_shared_ptr head; // in C++11: remove "atomic_" and use special // functions every time you touch