Very few things have ever been measured accurately to ten significant digits. How much did these numbers change run to run? How many measurements did you take? Were the caches warmed up similarly? Still, point taken.
The excessive "precision" is because I've just copy-pasted what the original bench printed.
As for "is this a reliable result", I believe I've performed diligence, appropriate for a HN comment, to make sure that this is not a completely random result. As I've said, I did not investigate this particular bit of code thoroughly. You are welcome to read the linked blog posts, which study the issue in depth.
As for "is this a reliable result", I believe I've performed diligence, appropriate for a HN comment, to make sure that this is not a completely random result. As I've said, I did not investigate this particular bit of code thoroughly. You are welcome to read the linked blog posts, which study the issue in depth.