Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 6 May 2009 20:13:02 +1000
From:      Peter Jeremy <peterjeremy@optushome.com.au>
To:        freebsd-current@freebsd.org
Subject:   hardclock() sources
Message-ID:  <20090506101301.GA17874@server.vk2pj.dyndns.org>

next in thread | raw e-mail | index | archive | help

--ZPt4rx8FFjLCG7dd
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Currently, FreeBSD uses the LAPIC timer to derive both hardclock()
and statclock()/profclock() if it exists, otherwise it uses the
8254 and RTC.  This presents a problem for using C3 - which stops
the LAPIC clock.

Some time ago, ariff@ produced a patch to work around a similar
problem with C1E on some AMD Turion CPUs:
http://people.freebsd.org/~ariff/misc/idlecpu_apic_5.diff

A better solution would seem to be to use a HPET periodic timer.  Is
there a particular reason why code to (at least optionally) support
HPET as the timer interrupt hasn't been written?  Looking back through
the archives, pkh@ has made disparaging remarks about the actual
usefulness of HPET (as against the Intel documentation) but I can't
find anything that actually talks about using it.

Note that supporting HPET as a timer source would provide at least
some of the infrastructure that will be needed for a future tickless
kernel.

--=20
Peter Jeremy

--ZPt4rx8FFjLCG7dd
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.11 (FreeBSD)

iEYEARECAAYFAkoBYq0ACgkQ/opHv/APuIfLngCfbgRt+478eQO1CaAtCPltykir
W2gAoI8Tga5coWvjU5X8wR7wDoRhy9x6
=rfRx
-----END PGP SIGNATURE-----

--ZPt4rx8FFjLCG7dd--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20090506101301.GA17874>