Skip to contentSkip to conversation

cpu governor ondemand issue

DISCUSSION Toralf Förster 3 participants 5 messages Oct 21, 2007 – Oct 29, 2007

lore
Share & tools
lore

Toralf Förster

author
#0 lore raw

I use the following one liner for regression tests of the cpufreq governor and friends:

tfoerste@n22 ~ $ time factor 819734028463158891

which usually needs 5.5 secs to complete at 1.7 GHz of my pentium M processor.

With the current git sources (commit cfa76f0 at Sat Oct 20 20:19:15 2007 -0700) instead this command needs 3 times longer if I run the "distributed.net client" as a background process with nice level 19 b/c the cpu frequency still stays at 600 MHz.
After stopping dnetc the cpu frequency governor ondemand works as expected again.

Tested at my ThinkPad T41 with stable Gentoo:

tfoerste@n22 ~ $ grep -e ^CONFIG_ACPI -e ^CONFIG_CPU_FREQ_GOV_ ~/devel/linux-2.6/.config CONFIG_ACPI=y
CONFIG_ACPI_SLEEP=y
CONFIG_ACPI_PROCFS=y
CONFIG_ACPI_PROC_EVENT=y
CONFIG_ACPI_AC=m
CONFIG_ACPI_BATTERY=m
CONFIG_ACPI_BUTTON=m
CONFIG_ACPI_VIDEO=m
CONFIG_ACPI_FAN=m
CONFIG_ACPI_PROCESSOR=m
CONFIG_ACPI_THERMAL=m
CONFIG_ACPI_BLACKLIST_YEAR=0
CONFIG_ACPI_EC=y
CONFIG_ACPI_POWER=y
CONFIG_ACPI_SYSTEM=y
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_POWERSAVE=m
CONFIG_CPU_FREQ_GOV_ONDEMAND=y

Signature · 4 lines
MfG/Sincerely

Toralf Förster
pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3
#1 lore raw
Quoted 26 lines · On Sun, 21 Oct 2007 13:16:14 +0200 Toralf Förster <[email protected]>…

I use the following one liner for regression tests of the cpufreq governor and friends:

tfoerste@n22 ~ $ time factor 819734028463158891

which usually needs 5.5 secs to complete at 1.7 GHz of my pentium M processor.

With the current git sources (commit cfa76f0 at Sat Oct 20 20:19:15 2007 -0700) instead this command needs 3 times longer if I run the "distributed.net client" as a background process with nice level 19 b/c the cpu frequency still stays at 600 MHz.
After stopping dnetc the cpu frequency governor ondemand works as expected again.

Tested at my ThinkPad T41 with stable Gentoo:

tfoerste@n22 ~ $ grep -e ^CONFIG_ACPI -e ^CONFIG_CPU_FREQ_GOV_ ~/devel/linux-2.6/.config CONFIG_ACPI=y
CONFIG_ACPI_SLEEP=y
CONFIG_ACPI_PROCFS=y
CONFIG_ACPI_PROC_EVENT=y
CONFIG_ACPI_AC=m
CONFIG_ACPI_BATTERY=m
CONFIG_ACPI_BUTTON=m
CONFIG_ACPI_VIDEO=m
CONFIG_ACPI_FAN=m
CONFIG_ACPI_PROCESSOR=m

Quoted 9 lines

CONFIG_ACPI_THERMAL=m
CONFIG_ACPI_BLACKLIST_YEAR=0
CONFIG_ACPI_EC=y
CONFIG_ACPI_POWER=y
CONFIG_ACPI_SYSTEM=y
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_POWERSAVE=m
CONFIG_CPU_FREQ_GOV_ONDEMAND=y

And 2.6.23 was OK?

Is this problem still present in Linus's current tree?

Thanks.

author
#2 lore raw

Am Montag, 29. Oktober 2007 schrieb Andrew Morton:

Quoted 38 lines · On Sun, 21 Oct 2007 13:16:14 +0200 Toralf Förster <[email protected]>…
Quoted 27 lines

I use the following one liner for regression tests of the cpufreq governor and friends:

tfoerste@n22 ~ $ time factor 819734028463158891

which usually needs 5.5 secs to complete at 1.7 GHz of my pentium M processor.

With the current git sources (commit cfa76f0 at Sat Oct 20 20:19:15 2007 -0700) instead this command needs 3 times longer if I run the "distributed.net client" as a background process with nice level 19 b/c the cpu frequency still stays at 600 MHz.
After stopping dnetc the cpu frequency governor ondemand works as expected again.

Tested at my ThinkPad T41 with stable Gentoo:

tfoerste@n22 ~ $ grep -e ^CONFIG_ACPI -e ^CONFIG_CPU_FREQ_GOV_ ~/devel/linux-2.6/.config CONFIG_ACPI=y
CONFIG_ACPI_SLEEP=y
CONFIG_ACPI_PROCFS=y
CONFIG_ACPI_PROC_EVENT=y
CONFIG_ACPI_AC=m
CONFIG_ACPI_BATTERY=m
CONFIG_ACPI_BUTTON=m
CONFIG_ACPI_VIDEO=m
CONFIG_ACPI_FAN=m
CONFIG_ACPI_PROCESSOR=m

Quoted 9 lines

CONFIG_ACPI_THERMAL=m
CONFIG_ACPI_BLACKLIST_YEAR=0
CONFIG_ACPI_EC=y
CONFIG_ACPI_POWER=y
CONFIG_ACPI_SYSTEM=y
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_POWERSAVE=m
CONFIG_CPU_FREQ_GOV_ONDEMAND=y

Quoted 6 lines

And 2.6.23 was OK?

Is this problem still present in Linus's current tree?

Thanks.

Hello,
with -rc1 the problem went away :-)

Signature · 4 lines
MfG/Sincerely

Toralf Förster
pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3
#3 lore raw
In reply to Andrew Morton
Quoted 4 lines · On Mon, 29 Oct 2007 00:02:43 -0700 Andrew Morton <[email protected]>…

And 2.6.23 was OK?

Is this problem still present in Linus's current tree?

and did you enable the "ignore nice levels" switch in sysfs ondemand directory?

Signature · 3 lines
If you want to reach me at my work email, use [email protected]
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org
author
#4 lore raw

Am Montag, 29. Oktober 2007 schrieb Arjan van de Ven:

Quoted 11 lines
Quoted 4 lines · On Mon, 29 Oct 2007 00:02:43 -0700 Andrew Morton <[email protected]>…

And 2.6.23 was OK?

Is this problem still present in Linus's current tree?

and did you enable the "ignore nice levels" switch in sysfs ondemand directory?

Oh no,
if you mean /sys/devices/system/cpu/cpu0/cpufreq/ondemand/ignore_nice_load I didn't changed anything, instead I just used the same .config file, rebooted into the new kernel and ran the test again.

Signature · 4 lines
MfG/Sincerely

Toralf Förster
pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3