Also, a yielded thread has absolutely no reason to be woken up with low latency, and this has nothing to do with the supposed scheduler quality. I'm not sure a kernel blindly prioritizing otherwise cpu-bound thread just because they yielded would give good results in the general case, esp out of microbenchmarks. If you want low wake up latency, you need a proper design, not mere reliance on magic that does not exist.

Plus highly contended critical sections should really really be avoided. So I'm not sure the author's benchmark even make any sense: if all of your workload is serialized, use a single thread.