From mboxrd@z Thu Jan 1 00:00:00 1970 From: Con Kolivas Subject: Re: [PATCH] i386 No Idle HZ aka dynticks v051205 Date: Mon, 5 Dec 2005 15:27:32 +1100 Message-ID: <200512051527.37791.kernel@kolivas.org> References: <200512051154.45500.kernel@kolivas.org> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============34879787940060658==" Return-path: In-Reply-To: <200512051154.45500.kernel@kolivas.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: ck-bounces@vds.kolivas.org Errors-To: ck-bounces@vds.kolivas.org To: ck@vds.kolivas.org Cc: vatsa@in.ibm.com, Tony Lindgren , Daniel Petrini , linux list , Dominik Brodowski , acpi-devel@lists.sourceforge.net List-Id: linux-acpi@vger.kernel.org --===============34879787940060658== Content-Type: multipart/signed; boundary="nextPart1869705.3TAPEQ9OTY"; protocol="application/pgp-signature"; micalg=pgp-sha1 Content-Transfer-Encoding: 7bit --nextPart1869705.3TAPEQ9OTY Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline On Monday 05 December 2005 11:54, Con Kolivas wrote: > Here is an updated rollup patch for current dynticks on i386. > > The main change to this version is the inclusion of Dominik's patches to > cpufreq ondemand, acpi c-states and bus mastering which should start maki= ng > the potential power saving features of dyntick a reality (thanks!). > One buildfix for !CONFIG_NO_IDLE_HZ as well. > > If you get strange stalls with this patch then almost certainly it is a > problem with dynticks and your apic so booting with the "noapic" option > should fix it. > > Split out patches, timertop and pmstats utilities and latest patch > available here: > http://ck.kolivas.org/patches/dyn-ticks/ > > FAQ: > What Hz should I use with dynticks in the config? > 1000 to realise the benefits of the power saving features and low latency. > > Should I enable timer statistics? > Only if you're planning on using the timertop utility to help you recogni= se > the biggest sources of timers currently in use to help you improve power > savings. Looks like this fix is needed if you are using cpufreq as modules. Cheers, Con =2D-- kernel/dyn-tick.c | 2 ++ 1 files changed, 2 insertions(+) Index: linux-2.6.15-rc5-dt/kernel/dyn-tick.c =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D =2D-- linux-2.6.15-rc5-dt.orig/kernel/dyn-tick.c +++ linux-2.6.15-rc5-dt/kernel/dyn-tick.c @@ -122,6 +122,8 @@ void dyn_early_reprogram(unsigned int de write_sequnlock_irqrestore(&xtime_lock, flags); } =20 +EXPORT_SYMBOL(dyn_early_reprogram); + void set_dyn_tick_limits(unsigned int max_skip, unsigned int min_skip) { if (max_skip > DYN_TICK_MAX_SKIP) --nextPart1869705.3TAPEQ9OTY Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQBDk8G5ZUg7+tp6mRURAtY0AJ4xPf3J8t4F75l3VTkbU3IdwaKeRwCePm2C mGIL4alGdalrQy3gQVqe1MI= =LXGp -----END PGP SIGNATURE----- --nextPart1869705.3TAPEQ9OTY-- --===============34879787940060658== Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline --===============34879787940060658==--