From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tony Lindgren Subject: Re: [RFC] [PATCH] ARM: Add clocksource driver for OMAP1 and OMAP2 Date: Fri, 1 Dec 2006 14:13:07 -0800 Message-ID: <20061201221306.GI23936@atomide.com> References: <4550A70B.7020609@gmail.com> <200611301114.09352.david-b@pacbell.net> <456F4025.5010403@gmail.com> <200611301322.35918.david-b@pacbell.net> <45707299.3080207@gmail.com> <20061201190317.B12841DFAE4@adsl-69-226-248-13.dsl.pltn13.pacbell.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20061201190317.B12841DFAE4@adsl-69-226-248-13.dsl.pltn13.pacbell.net> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: linux-omap-open-source-bounces@linux.omap.com Errors-To: linux-omap-open-source-bounces@linux.omap.com To: David Brownell Cc: linux-omap-open-source@linux.omap.com List-Id: linux-omap@vger.kernel.org * David Brownell [061201 11:11]: > > > + clocksource_khz2mult(32768, clocksource_32k.shift); > > > > Using clocksource_hz2mult instead (you see, hz vs. khz) does > > the trick here ;) > > Ah, the old "off by one (letter)" trick. I knew it had to be > something like that ... thanks! :) > > Evidently you agree that using a second timer is appropriate. > > However, I'm not sure that a better solution might not be to > register the 32k timer clocksource by default (except on 1510), > since it seems cheap (and "always available"). Clocksource > logic can select a higher precision one if it's available. Yeah, I think the 32k sync timer for clock source would be best. On omap2, the 32k gptimer1 is needed for timer interrupts to produce wake-up events. > Even with as many non-32k timer options as OMAP has (dmtimer for > both OMAP1 and OMAP2, MPU timer for OMAP1, 32k), I could imagine > usage conflicts, as reported this AM on l-a-k for ixp4xx. And > the only really general way around those seems to be a Kconfig > option for which clocksource to use (other than 32k sync timer). > > Comments? Using the 32k sync timer for the clocksource is best from PM point of view as it always runs. I'd rather go for that by default, and then have the optional mpu timer clocksource available as a Kconfig option. Regards, Tony