Interesting when you start to compare with C++ stdlib which only has shared_ptr, being the equivalent to Arc, but no single-threaded equivalent Rc. Usually the

Interesting when you start to compare with C++ stdlib which only has shared_ptr, being the equivalent to Arc, but no single-threaded equivalent Rc. Usually the argument you hear is that the performance difference is negligible. One could also argue that you shouldn't use any kind of refcounter in a tight loop where it can affect performance as much as it did there. Refactoring the code a bit to allow passing by reference there would likely help even more. Still cool to see that just changing to Rc gave that much gain.