The discussion of running Python without a GC reminded me that reference counting doesn’t have to be slow at all! This algorithm gets to within 10% of highly tuned generational collectors, but is fundamentally reference counting.
The discussion of running Python without a GC reminded me that reference counting doesn’t have to be slow at all! This algorithm gets to within 10% of highly tuned generational collectors, but is fundamentally reference counting.
Highly recommend this article! IIRC, they make a pretty good case that deferred counts behave similarly to generations.