Date: Sun, 26 Aug 2001 16:06:57 +1000 (EST) From: Bruce Evans <bde@zeta.org.au> To: Mike Smith <msmith@FreeBSD.ORG> Cc: Thierry Herbelot <thierry@herbelot.com>, David Hill <david@phobia.ms>, <current@FreeBSD.ORG>, <asmodai@FreeBSD.ORG> Subject: Re: options HZ [MFC for pr=28143] Message-ID: <20010826155534.E15196-100000@besplex.bde.org> In-Reply-To: <200108260115.f7Q1FNn22234@mass.dis.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 25 Aug 2001, Mike Smith wrote: > It should be a tunable, not a compile-time option. > > > David Hill wrote: > > > > > > Hello - > > > Could someone please document "options HZ" into LINT?. I found it whil > > e > > > reading the dummynet(4) manpage. It must remain a compile-time option for those of us who don't believe in tunables. Notes: - Making HZ non-constant (almost everything uses `hz' instead of HZ) mainly pessimizes time-critical code that computes hz/CONST, except: - on alphas, the HZ option has no effect. The initial value of hz = HZ is not used except to compute wrong values for `tick' and `tickadj' (`tick' is fixed up later but `tickadj' remains wrong). `hz' is initialized to hwprb->rpb_intr_freq >> 12 in alpha/machdep.c. So, hz needs to be non-constant and the HZ option and tunables just get in the way for alphas and for other arches where the hardclock frequency is not programmable. Bruce To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010826155534.E15196-100000>