Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754953AbbKMTyI (ORCPT ); Fri, 13 Nov 2015 14:54:08 -0500 Received: from mga01.intel.com ([192.55.52.88]:24041 "EHLO mga01.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754885AbbKMTyE (ORCPT ); Fri, 13 Nov 2015 14:54:04 -0500 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.20,289,1444719600"; d="scan'208";a="684847545" From: Jacob Pan To: Peter Zijlstra , Ingo Molnar , Thomas Gleixner , John Stultz , LKML Cc: Arjan van de Ven , Srinivas Pandruvada , Len Brown , Rafael Wysocki , Eduardo Valentin , Paul Turner , Jacob Pan Subject: [PATCH 0/4] CFS idle injection Date: Fri, 13 Nov 2015 11:53:03 -0800 Message-Id: <1447444387-23525-1-git-send-email-jacob.jun.pan@linux.intel.com> X-Mailer: git-send-email 1.9.1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2628 Lines: 66 We are entering a very power and thermal constrained environment. Often we have more horsepower than we can use due to these limits. But on the other side we all demand performance when needed. The reserved performance headroom does not come free. To conserve energy, more and more SoC blocks can be power gated at runtime. However, randomly scheduled idle time on individual CPU may not result in good power saving in that the common circuits such as memory controller can only be in the low power state when all cores are in idle at the same time. Frequency-Voltage scaling presents a good solution but its efficiency is limited to certain range. In general, only synchronized idle will allow SoC to enter the deepest power states. For most modern processors, deep idle power is nearly negligible to the peak running power. This implies if we can duty cycle the CPU between running and the deepest idle state, we can scale performance and power almost linearly. Combined with the most efficient frequency point, idle injection presents a way to cap power efficiently. Intel powerclamp driver was introduced a while ago to address the problem but is broken in the sense of turning off idle ticks in the forced idle period. https://lkml.org/lkml/2014/12/18/369 It was suggested to replace the current kthread play idle loop with a timer based runqueue throttling scheme. I finally got around to implement this and code looks much simpler and more effective. Test results were presented at LinuxCon where data/graph can be seen from slides #18 and later. http://events.linuxfoundation.org/sites/events/files/slides/LinuxCon_Japan_2015_idle_injection1_0.pdf RFC discussions are here. https://lkml.org/lkml/2015/11/2/756 Thanks, Jacob Jacob Pan (4): ktime: add a roundup function timer: relax tick stop in idle entry sched: introduce synchronized idle injection sched: add trace event for idle injection include/linux/ktime.h | 10 ++ include/linux/sched.h | 16 ++ include/linux/sched/sysctl.h | 5 + include/trace/events/sched.h | 25 +++ init/Kconfig | 10 ++ kernel/sched/fair.c | 356 ++++++++++++++++++++++++++++++++++++++++++- kernel/sched/sched.h | 54 ++++++- kernel/sysctl.c | 21 +++ kernel/time/tick-sched.c | 2 +- 9 files changed, 493 insertions(+), 6 deletions(-) -- 1.9.1 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/