From mboxrd@z Thu Jan  1 00:00:00 1970
From: tglx@linutronix.de (Thomas Gleixner)
Date: Thu, 8 Sep 2011 11:34:59 +0200 (CEST)
Subject: [PATCH 3/4] mach-integrator: modernize clock event
 registration
In-Reply-To: <CACRpkdZ0HgvhgFU9-4f7QtWT3aJuZhY8QQB+oEW0mEy+Uzuh=Q@mail.gmail.com>
References: <1315384260-27404-1-git-send-email-linus.walleij@stericsson.com>
	<alpine.LFD.2.02.1109071202000.2723@ionos>
	<20110908085502.GB14953@n2100.arm.linux.org.uk>
	<CACRpkdZ0HgvhgFU9-4f7QtWT3aJuZhY8QQB+oEW0mEy+Uzuh=Q@mail.gmail.com>
Message-ID: <alpine.LFD.2.02.1109081133530.2723@ionos>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Thu, 8 Sep 2011, Linus Walleij wrote:

> On Thu, Sep 8, 2011 at 10:55 AM, Russell King - ARM Linux
> <linux@arm.linux.org.uk> wrote:
> >> Acked-by: Thomas Gleixner <tglx@linutronix.de>
> 
> > My understanding is that periodic timers are not setup by the generic
> > clock event code, and so we do need to keep the 'timer_reload' stuff
> > around.
> 
> Aha. Yes maybe I've not tested that mode well enough, I'll
> have a closer look.
> 
> An alteranative is to retire the periodic mode and only
> support oneshot, like the plat-nomadik/timer.c currently
> does. My impression is that oneshot will always work fine,
> and doing this leaves all details up to the timekeeping
> core so we don't need to deal with it, but I may be wrong.
> Ideas?

It does, but for a pure periodic setup (nohz=off, highres=off) you add
the overhead of reprogramming the device for each tick.

Thanks,

	tglx