> Date: Mon, 15 Jan 2018 01:33:41 +0900 > From: "Naruaki.Etomi" <nullnil...@gmail.com> > > I propose to fix buggy PCC.
Cool, thanks for working on this! I don't have time to review it right now, but the main point that is broken in many of our per-CPU timecounters is that the timecounter API is designed to provide a _global_ clock, so it gets confused if you only present a per-CPU clock. Of course you can always turn a per-CPU clock into a global clock by maintaining the most recent global timestamp the system has seen using atomic operations or a lock. One of these days we should add a per-CPU timecounter API and a generic way to make a global timecounter out of it -- but we don't have that yet.