What is "extra timer interrupt" in powertop output?

What is "extra timer interrupt" in powertop output?

[Andrey Borzenkov]

[-- Attachment #1: Type: text/plain, Size: 1539 bytes --]

After disabling CD-ROM polling the next top-100 winner is "extra timer 
interrupt":

Cn                Avg residency       P-states (frequencies)
C0 (cpu running)        ( 4.2%)          750 Mhz     0.0%
polling           0.0ms ( 0.0%)          500 Mhz   100.0%
C1                0.0ms ( 0.0%)
C2                0.9ms (95.8%)


Wakeups-from-idle per second : 1101.2   interval: 5.0s
Power usage (ACPI estimate): 11.7W (2.2 hours)

Top causes for wakeups:
  31.6% (  8.4)       <interrupt> : extra timer interrupt
  15.0% (  4.0)   hald-addon-acpi : schedule_timeout (process_timeout)

I assume, this is what is simply called "timer":
{pts/1}:~%> cat /proc/interrupts
           CPU0
  0:    9060975    XT-PIC-XT        timer
  1:      18742    XT-PIC-XT        i8042
  2:          0    XT-PIC-XT        cascade
  3:          2    XT-PIC-XT
  4:          2    XT-PIC-XT
  5:          3    XT-PIC-XT
  7:          5    XT-PIC-XT        irda0
  8:         11    XT-PIC-XT        rtc0
  9:       3376    XT-PIC-XT        acpi
 10:          2    XT-PIC-XT
 11:     239251    XT-PIC-XT        ohci_hcd:usb1, yenta, yenta, yenta, ALI 
5451, pcmcia0.0
 12:     621206    XT-PIC-XT        i8042
 14:     225993    XT-PIC-XT        ide0
 15:     315463    XT-PIC-XT        ide1
NMI:          0   Non-maskable interrupts
ERR:          0

Any way to find out why there are so many of them? System is using dynamic 
ticks:
{pts/1}:~/src/linux-git%> grep NO_HZ /boot/config
CONFIG_NO_HZ=y

TIA

-andrey



[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: What is "extra timer interrupt" in powertop output?

[Alan Jenkins]

Andrey Borzenkov wrote:
> After disabling CD-ROM polling the next top-100 winner is "extra timer
> interrupt":
>
> Cn Avg residency P-states (frequencies)
> C0 (cpu running) ( 4.2%) 750 Mhz 0.0%
> polling 0.0ms ( 0.0%) 500 Mhz 100.0%
> C1 0.0ms ( 0.0%)
> C2 0.9ms (95.8%)
>
>
> Wakeups-from-idle per second : 1101.2 interval: 5.0s
> Power usage (ACPI estimate): 11.7W (2.2 hours)
>
> Top causes for wakeups:
> 31.6% ( 8.4) : extra timer interrupt
> 15.0% ( 4.0) hald-addon-acpi : schedule_timeout (process_timeout)
>
> I assume, this is what is simply called "timer":
> {pts/1}:~%> cat /proc/interrupts
> CPU0
> 0: 9060975 XT-PIC-XT timer

>
> Any way to find out why there are so many of them? System is using dynamic
> ticks:
> {pts/1}:~/src/linux-git%> grep NO_HZ /boot/config
> CONFIG_NO_HZ=y
>

One possible reason is the hardware timer used cannot delay for long
periods, so the kernel has to wake up more often than necessary.  Um. 
This doesn't seem to be explained well in docs.

The question is do you have HPET (good hardware timer) enabled? 
Enabling HPET is supposed to be one of the things powertop suggests. 
Sometimes it is a BIOS option, or may require a kernel option, or may
not be supported on your motherboard.

Alan