Ubuntu EeePC-optimized Kernel

No detailed statistics in PowerTOP on 2.6.27-7-eeepc-lean

Reported by Evgueny on 2008-10-27
8
Affects Status Importance Assigned to Milestone
Ubuntu EeePC Kernel
Undecided
Unassigned

Bug Description

Asus Eee PC 901

Intel PowerTOP v.1.9 on kernel 2.6.27-7-eeepc-lean reports:

No detailed statistics available; please enable the CONFIG_TIMER_STATS kernel option
This option is located in the Kernel Debugging section of menuconfig
(which is CONFIG_DEBUG_KERNEL=y in the config file)
Note: this is only available in 2.6.21 and later kernels

It would be nice to get back detailed power profiling in case there are no major setbacks from using debug option in kernel.

Peter Gantner (eeepc) wrote :

bump, seconded.

Powertop is a *very* handy tool, esp. for a laptop, and setting that kernel option carries very low risk of something stopping to work and has the benefit of changing the PT output from the stuff quoted above to something like this (from another machine), which is much more useful:

[...]
Top causes for wakeups:
  35.2% (206.2) java : futex_wait (hrtimer_wakeup)
  19.9% (116.4) <interrupt> : ide0
  13.6% ( 79.4) <kernel core> : ide_do_rw_disk (ledtrig_ide_timerfunc)
  13.5% ( 79.0) java : schedule_timeout (process_timeout)
  10.2% ( 60.0) <interrupt> : ATI IXP, fglrx[0]@PCI:1:5:0
   2.8% ( 16.2) <interrupt> : b43
   1.6% ( 9.6) porthole : schedule_timeout (process_timeout)
   0.6% ( 3.8) <interrupt> : ohci_hcd:usb1, ohci_hcd:usb2
   0.6% ( 3.8) USB device 2-3 : PS/2+USB Mouse ()
   0.3% ( 2.0) kded : futex_wait (hrtimer_wakeup)
[...]

PowerTop still complains of missing CONFIG_TIMER_STATS using 2.6.27-11-eeepc-lean, but not when running 2.6.27-11-eeepc or 2.6.27-11-generic.

When using 2.6.27-11-eeepc-lean PowerTop also sugests disabling CONFIG_IRQBALANCE, which it doesn't do when running 2.6.27-11-eeepc or 2.6.27-11-generic, but I don't know enough to say if PowerTop is right on that one, but you might want to investigate the issue.

Peter Gantner (eeepc) wrote :

About irqbalancing:

Generally it is suggested to balance IRQs across all available CPUs for performance reasons.

I think the reason powertop suggests turning it off is when you have two real CPUs or cores, balancing IRQs can prohibit an otherwise idle CPU from entering deep powersaving states.

Not sure what the situation is in HT setups like the Eee (Atom) but I don't think it can matter there power-saving-wise.

Sitsofe Wheeler (sitsofe) wrote :

Powertop is saying "don't let the kernel do irqbalancing by itself - it won't do as good a job as leaving it to a userspace tool". See http://www.lesswatts.org/tips/cpu.php for further details.

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers