> So the code in question is pure garbage. You can't do spinlocks like that. Or rather, you very much can do them like that, and when you do that you are measur

> So the code in question is pure garbage. You can't do spinlocks like that. Or rather, you very much can do them like that, and when you do that you are measuring random latencies and getting nonsensical values, because what you are measuring is "I have a lot of busywork, where all the processes are CPU-bound, and I'm measuring random points of how long the scheduler kept the process in place". > And then you write a blog-post blamings others, not understanding that it's your incorrect code that is garbage, and is giving random garbage values. > And then you test different schedulers, and you get different random values that you think are interesting, because you think they show something cool about the schedulers. Tldr: They got probably lucky on Windows. Conceivably similar code might have worked on Linux and then had issues on Windows.