Date: Sun, 10 Oct 2021 00:21:39 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: Warner Losh <imp@bsdimp.com>, Sebastian Huber <sebastian.huber@embedded-brains.de>, FreeBSD Hackers <freebsd-hackers@freebsd.org> Subject: Re: Why was the timehands_count sysctl added? Message-ID: <YWIH426AQNC4qoY5@kib.kiev.ua> In-Reply-To: <202110092107.199L7T4j059128@critter.freebsd.dk> References: <2d1d2a6d-ec6b-7f52-8af3-09a833c52820@embedded-brains.de> <YWHzoTpcZTQcHVUw@kib.kiev.ua> <CANCZdfr7_jb07%2BAft_uo2F8L9hDr9iABaDkANioJhi-kQXQeoQ@mail.gmail.com> <202110092107.199L7T4j059128@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Oct 09, 2021 at 09:07:29PM +0000, Poul-Henning Kamp wrote: > -------- > Warner Losh writes: > > > > To allow for experimentation, and to satisfy some requests where people > > > wanted to have more that 2 timehands. > > > > When would someone want that? What's the use case? > > The reason there were originally 10 timehands was that latency in > the early SMP kernels was ... ehh ... variable ... and some of the > time-counters rolled over quite fast compared to that. > > I really hope no relevant current hardware has that problem. The current algorithm to read timehands is resilient to the wrap-out of the current hand. You really need to experience enourmous delays in the reader loop to make it lock-step with tc_windup() updates, in which case it could indeed be better to have more than two timehands. I believe it was Ian who reported that 16 timehands worked better for him than 2.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?YWIH426AQNC4qoY5>