From mboxrd@z Thu Jan 1 00:00:00 1970 From: tglx@linutronix.de (Thomas Gleixner) Date: Tue, 22 Sep 2015 22:53:14 +0200 (CEST) Subject: [PATCH v2 7/7] ARM: smp: Add runtime PM support for CPU hotplug In-Reply-To: <20150922173236.GB19183@linaro.org> References: <2657565.lxWtPmdjyp@vostro.rjw.lan> <55ED9328.2010501@ti.com> <2355864.I8qZFusWfm@vostro.rjw.lan> <55EE9AA1.8020200@ti.com> <7h613kmuuv.fsf@deeprootsystems.com> <55F162FD.4050401@ti.com> <20150922173236.GB19183@linaro.org> Message-ID: To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Tue, 22 Sep 2015, Lina Iyer wrote: > Talking to John Stultz at Linaro Connect: Is cpuidle relevant in > -RT kernel? I dont know much about -RT. I thought this might be point > that we should consider. There are definitely RT systems out there where power consumption is a concern. > As it stands today, on a 800 Mhz ARM quad core, I am seeing a latency of > 50-70 usec for the additional runtime PM in the cpuidle. Ofcourse, there > is a definite need to optimize and there are opportunities to do that. RT is not about being fast. RT is about being deterministic. So if a power sensitive RT system can afford the extra latencies induced by PM it will tolerate them. Of course, optimizing that is not a bad thing at all :) Thanks, tglx