Load balancing tick
Posted: 11 Nov 2010, 17:11
Greeting,
using a NP220 running with lucid 10.04.
I'm not using HAL and therefore have no laptop-mode.
As shown by powertop, the process [kernel scheduler] Load balancing tick is the main reason for waking up on my little box. On a standard user setting (firefox /w several tabs, nautilus, terminal, wlan) it typically rivals with firefox for the first place, sometimes about the double of amount (of ticks?), sometimes even more.
Example output of powertop:
As far as I understand, the load balancing tick is a kernel issue cause of traditional linux kernels and can be made tickless to save some energy (-> http://www.lesswatts.org/projects/tickless/).
Also there is/was a bug in newer kernels (2.6.34) concerning this (-> https://groups.google.com/group/zen_ker ... f1a3a7b318).
No I ask myself if in your kernel we all use there may be the same or a similar bug, and if it would make sense to make the kernel tickless by default (if it isn't already).
If you think that is of importance, I would gladly help e.g. by compiling and testing such a kernel. But you needed to point me to the config and any obscure resources/patches of the recent ppa kernel.
using a NP220 running with lucid 10.04.
Code: Select all
peter@rattir7:~$ uname -a
Linux rattir7 2.6.32-25-generic #46~ppa1~loms~lucid-Ubuntu SMP Fri Oct 22 23:33:19 UTC 2010 i686 GNU/Linux
As shown by powertop, the process [kernel scheduler] Load balancing tick is the main reason for waking up on my little box. On a standard user setting (firefox /w several tabs, nautilus, terminal, wlan) it typically rivals with firefox for the first place, sometimes about the double of amount (of ticks?), sometimes even more.
Example output of powertop:
Code: Select all
peter@rattir7:~$ env LANG=en_EN.UTF8 sudo powertop -d
PowerTOP 1.12 (C) 2007, 2008 Intel Corporation
Collecting data for 15 seconds
Your CPU supports the following C-states : C1 C2 C4
Your BIOS reports the following C-states : C1 C2 C4
Cn Avg residency
C0 (cpu running) ( 6.8%)
polling 0.0ms ( 0.0%)
C1 mwait 0.2ms ( 0.1%)
C2 mwait 0.8ms ( 1.2%)
C4 mwait 5.6ms (91.9%)
P-states (frequencies)
1.67 Ghz 3.5%
1333 Mhz 0.3%
1000 Mhz 96.2%
Disk accesses:
Wakeups-from-idle per second : 182.2 interval: 15.0s
Power usage (ACPI estimate): 6.1W (5.0 hours)
Top causes for wakeups:
30.6% ( 70.8) firefox-bin
26.9% ( 62.1) [kernel scheduler] Load balancing tick
11.9% ( 27.6) [uhci_hcd:usb5, eth1] <interrupt>
11.4% ( 26.4) [extra timer interrupt]
4.3% ( 10.0) xbindkeys
3.6% ( 8.4) [kernel core] hrtimer_start (tick_sched_timer)
2.6% ( 6.1) [kernel core] add_timer (wl_timer)
Also there is/was a bug in newer kernels (2.6.34) concerning this (-> https://groups.google.com/group/zen_ker ... f1a3a7b318).
No I ask myself if in your kernel we all use there may be the same or a similar bug, and if it would make sense to make the kernel tickless by default (if it isn't already).
If you think that is of importance, I would gladly help e.g. by compiling and testing such a kernel. But you needed to point me to the config and any obscure resources/patches of the recent ppa kernel.