Date: Fri, 18 Jun 2021 09:49:46 -0700 From: John Baldwin <jhb@FreeBSD.org> To: Eugene Grosbein <eugen@grosbein.net>, Warner Losh <imp@FreeBSD.org>, src-committers@FreeBSD.org, dev-commits-src-all@FreeBSD.org, dev-commits-src-main@FreeBSD.org Subject: Re: git: 4c0bc591466f - main - man9: add hz(9) and hardclock(9) Message-ID: <c8676c23-f2f3-6430-3a48-8696485b591c@FreeBSD.org> In-Reply-To: <7c4c7acc-1fe4-28e4-4e1c-c08912eb0f99@grosbein.net> References: <202106181443.15IEhn4t010735@gitrepo.freebsd.org> <10df4ab6-bee0-58f5-7d42-92850d526e01@FreeBSD.org> <7c4c7acc-1fe4-28e4-4e1c-c08912eb0f99@grosbein.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 6/18/21 8:32 AM, Eugene Grosbein wrote: > 18.06.2021 22:18, John Baldwin wrote: > >> OTOH, it's >> also true that there's no real reason for anything outside of the actual timer >> code to use stathz (or even profhz) unlike 'hz' which is still used to set >> timeout tick values. > > Not agreed: how do I get reliable per-CPU load stats in userland without sysctl kern.cp_times > that exports incrementing raw "stathz-tick" counters? I need them to draw per-CPU graphs. Hmm, I guess userland needs the resolution of the values in the sysctls, yes. To be clear, I was not at all saying that cp_times should be removed, only that very little code in the kernel needs to use the literal C symbol 'stathz' compared to the C symbol 'hz'. -- John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?c8676c23-f2f3-6430-3a48-8696485b591c>