> But the original article is completely, laughably wrong about one thing: an atomic increment or decrement is a remarkably slow operation on modern hardware, second only to pointer chasing.

Very true. No matter how one looks at it, GC imposes a non-trivial overhead. If I remember correctly, wasn’t there a study that demonstrated an average Swift application spending 40% of its time on RC?

But at the same time hardware can be improved. Apple CPUs are an obvious example. Uncontended atomics are extremely cheap and the memory prefetcher can detect and prefetch pointer chases.