> I want the equivalent of atomic<shared_ptr<T>>, but lock-free. You can do that without transactions with a lock cmpxchg16b instruction, or with a more compli

> I want the equivalent of atomic<shared_ptr<T>>, but lock-free. You can do that without transactions with a lock cmpxchg16b instruction, or with a more complicated algorithm like this: https://github.com/facebook/folly/blob/3de8f357eecaff89e3847fdb4f090767c2d42e81/folly/concurrency/AtomicSharedPtr.h https://github.com/facebook/folly/blob/3de8f357eecaff89e3847...