Messages in this thread | | | Subject | [RFC PATCH v1 0/2] sched: unified sched_powersavings tunables | From | Vaidyanathan Srinivasan <> | Date | Mon, 16 Jan 2012 21:52:26 +0530 |
| |
Hi,
I have created the following RFC patch based on the recent discussions and consensus on simplifying the power aware scheduler in the kernel.
Ref: LWN: Rethinking power-aware scheduling http://lwn.net/Articles/474915/
The goal of the unified tunable are as follows:
* Simple and single tunable for different (all) topology * Good default powersavings for kernel * Potential to map this setting to other subsystems like cpufreq and cpuidle
What this patch does: (first step)
* Create a single sched_powersavings tunable in sysfs * Enable current sched_mc and sched_smt features based on value of this single tunable
What this patch is yet to do:
* Tune the default power savings to pack packages only till a threshold of say 50% * Add notifiers to change setting on battery/AC transitions * Feed the thresholds from arch specific code so that different archs can do 'optimal' packing based on topology * May be move this to /sys/device/system/powersavings and add additional platform tunables like x86_energy_perf_policy?
This RFC has only x86 changes and has been tested on dual-socket, quad-core,HT configuration.
Please let me know your comments and feedback.
Thanks, Vaidy
---
Vaidyanathan Srinivasan (2): sched: unified sched_powersavings sysfs tunable sched: fix group_capacity for thread level consolidation
arch/x86/Kconfig | 20 ++++-------- arch/x86/kernel/smpboot.c | 2 + block/blk.h | 11 ++++--- drivers/base/cpu.c | 2 + include/linux/sched.h | 29 +++++++++-------- include/linux/topology.h | 9 +---- kernel/sched/core.c | 75 +++++++++++---------------------------------- kernel/sched/fair.c | 38 ++++++++++++++++------- 8 files changed, 77 insertions(+), 109 deletions(-)
| |