Boost’s shared_ptr up to 10× slower than OCaml’s garbage collection:

Modern garbage collector beats reference counting; news at 11. One interesting paragraph:

Note also the new “stack” section that is the first C to beat OCaml, albeit cheating by exploiting the fact that this implementation of this benchmark always happens to allocate and free in FIFO order. So this approach is problem-specific and cannot be used as a general-purpose allocator. However, it is worth noting that the only allocation strategy to beat C also (like OCaml’s nursery) exploits the ability to collect many freed values in constant time by resetting the stack pointer rather than using repeated pops. Therefore, the only allocation strategies likely to approach OCaml’s performance on these kinds of problems are those that exploit sparsity. For example, by manipulating free lists in run-length encoded form.