Page: 1, 2, 3  Next

braveheartleo
Status: Interested
Joined: 23 Jan 2012
Posts: 29
Back to top
Posted: Feb 2, 12, 17:50    Is cpufreq scaling missing in liquorix?
The ability to tweak cpufreq scaling via:
:: Code ::
/sys/devices/system/cpu/cpu*/cpufreq/scaling_*
seems to be missing from liquorix kernels.

Is liquorix patching this ability somehow in some way that still enables frequency scaling in a procie? Powertop doesn't report Frequency stats while running on a liquorix kernel. However, it reports using 'ondemand' cpufreq governor.
damentz
Status: Assistant
Joined: 09 Sep 2008
Posts: 530
Back to top
Posted: Feb 2, 12, 20:02    
Honestly, that sounds like a bug in powertop. Ever since they released the new version, I haven't been able to get _any_ stats at all. I haven't changed my configuration any, and it seems a bit suspicious that powertop would want me to enable high overhead options like CPU_FREQ_STAT_DETAILS that offer nothing useful until you begin debugging your scaling stats.

I'll look online later to see what their new requirements are to getting powertop working.
braveheartleo
Status: Interested
Joined: 23 Jan 2012
Posts: 29
Back to top
Posted: Feb 2, 12, 20:57    
The comparison was made against a stock kernel that came with the distro. Using the stock kernel I can access cpufreq scaling params via:
:: Code ::
/sys/devices/system/cpu/cpu*/cpufreq/scaling_*

However, in liquorix, this path is non-existent.

Another observation is that I use a monitoring utility, conky, that reports the current frequency under different system loads. Using the liquorix kernel this frequency remains pinned at the CPU's maximum frequency, which lead me to believe that scaling is not implemented in liquorix, or is implemented in a different way that utilities such as conky and powertop doesn't seem to support.
braveheartleo
Status: Interested
Joined: 23 Jan 2012
Posts: 29
Back to top
Posted: Feb 2, 12, 21:02    
:: damentz wrote ::
Honestly, that sounds like a bug in powertop. Ever since they released the new version, I haven't been able to get _any_ stats at all.

Using a stock kernel Frequency stats in powertop is available. However, the same is not in liquorix. Powertop is at version 1.97-2 on Xubuntu Linux 11.10 32-bit.
damentz
Status: Assistant
Joined: 09 Sep 2008
Posts: 530
Back to top
Posted: Feb 4, 12, 12:48    
According to a bug 834725 on launchpad, powertop 2.0 beta uses deprecated API that's supposed to have already been removed in Linux 3.2:

:: Quote ::
Deprecated power event trace API, to be removed (EVENT_POWER_TRACING_DEPRECATED)

CONFIG_EVENT_POWER_TRACING_DEPRECATED:

Provides old power event types:
C-state/idle accounting events:
power:power_start
power:power_end
and old cpufreq accounting event:
power:power_frequency
This is for userspace compatibility
and will vanish after 5 kernel iterations,
namely 3.1.


I'll enable this option for the next liquorix package, but don't be surprised if powertop stops working in 3.3.
braveheartleo
Status: Interested
Joined: 23 Jan 2012
Posts: 29
Back to top
Posted: Feb 4, 12, 13:01    
Good find on the powertop issue. Thank you. Could the hint indicated there also be the same with why conky is not reporting frequency changes under different system loads while using liquorix kernel?

However, that still doesn't answer why the cpufreq scaling params are missing from the path I previously indicated, or perhaps they have moved elsewhere in liquroix?
damentz
Status: Assistant
Joined: 09 Sep 2008
Posts: 530
Back to top
Posted: Feb 6, 12, 15:35    
Here's the last 10 lines of listing the path you gave me:

:: Code ::
$ ls -lh /sys/devices/system/cpu/cpu*/cpufreq/scaling_* | tail
-rw-r--r-- 1 root root 4.0K Feb  5 20:41 /sys/devices/system/cpu/cpu2/cpufreq/scaling_min_freq
-rw-r--r-- 1 root root 4.0K Feb  6 17:32 /sys/devices/system/cpu/cpu2/cpufreq/scaling_setspeed
-r--r--r-- 1 root root 4.0K Feb  6 17:32 /sys/devices/system/cpu/cpu3/cpufreq/scaling_available_frequencies
-r--r--r-- 1 root root 4.0K Feb  6 17:32 /sys/devices/system/cpu/cpu3/cpufreq/scaling_available_governors
-r--r--r-- 1 root root 4.0K Feb  6 17:32 /sys/devices/system/cpu/cpu3/cpufreq/scaling_cur_freq
-r--r--r-- 1 root root 4.0K Feb  6 17:32 /sys/devices/system/cpu/cpu3/cpufreq/scaling_driver
-rw-r--r-- 1 root root 4.0K Feb  5 20:41 /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor
-rw-r--r-- 1 root root 4.0K Feb  5 20:41 /sys/devices/system/cpu/cpu3/cpufreq/scaling_max_freq
-rw-r--r-- 1 root root 4.0K Feb  5 20:41 /sys/devices/system/cpu/cpu3/cpufreq/scaling_min_freq
-rw-r--r-- 1 root root 4.0K Feb  6 17:32 /sys/devices/system/cpu/cpu3/cpufreq/scaling_setspeed


Perhaps you're not running cpufreq to enable power management on your cpu. Can you run inxi -SMC so we can figure out what kind of system we're dealing with?
braveheartleo
Status: Interested
Joined: 23 Jan 2012
Posts: 29
Back to top
Posted: Feb 6, 12, 21:15    
Here's inxi -SMC on my system:
:: Code ::
System:    Host: host Kernel: 3.2.0-4.dmz.1-liquorix-686 i686 (32 bit) Desktop: Xfce 4.8.3 Distro: Ubuntu 11.10 oneiric
Machine:   Mobo: ASUSTeK model: P5PE-VM version: Rev 1.03G serial: 0
           Bios: American Megatrends version: 1501 date: 12/26/2007
CPU:       Single core Intel Pentium 4 CPU (-HT-) cache: 2048 KB flags: (lm nx sse sse2 sse3)
           Clock Speeds: 1: 3000.036 MHz 2: 3000.036 MHz

The paths exist in Ubuntu Kernel Linux 3.0.0.0-15-generic, but not in Liquorix. Could the differences between Ubuntu and Debian, which is the target audience of Liquorix kernels, have bearing upon this issue?
damentz
Status: Assistant
Joined: 09 Sep 2008
Posts: 530
Back to top
Posted: Feb 6, 12, 21:40    
It could be that the Pentium 4 had fake speedstep, also known as clock modulation. This wasn't really helpful since it just divided the amount time the CPU spent working evenly between the time it was not. For instance, if your maximum speed was 1ghz, and it clocks you down to 750mhz, you're actually idling 25% of the time.

I could enable speedstep clock modulation, but it honestly doesn't help with power, especially on laptops. If anything, it'll make your throughput slower, since you're getting no real measurable reduction in power, and your system is spending time deciding how much to divide your CPU power between idling and real work.

Let me know what CPU family you have. If it so happens that it's a version with enhanced speedstep, we can look further.
braveheartleo
Status: Interested
Joined: 23 Jan 2012
Posts: 29
Back to top
Posted: Feb 6, 12, 21:56    
Here's /proc/cpuinfo:
:: Code ::
processor   : 0
vendor_id   : GenuineIntel
cpu family   : 15
model      : 6
model name   : Intel(R) Pentium(R) 4 CPU 3.00GHz
stepping   : 5
cpu MHz      : 2400.000
cache size   : 2048 KB
physical id   : 0
siblings   : 2
core id      : 0
cpu cores   : 1
apicid      : 0
initial apicid   : 0
fdiv_bug   : no
hlt_bug      : no
f00f_bug   : no
coma_bug   : no
fpu      : yes
fpu_exception   : yes
cpuid level   : 6
wp      : yes
flags      : fpu vme de pse tsc msr pae mce cx8 apic mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe nx lm constant_tsc pebs bts pni dtes64 monitor ds_cpl est tm2 cid cx16 xtpr pdcm lahf_lm
bogomips   : 6000.00
clflush size   : 64
cache_alignment   : 128
address sizes   : 36 bits physical, 48 bits virtual
power management:

This cpu has Enhanced Intel SpeedStep Tech (EIST) setting in the BIOS.

I haven't read on clock modulation yet. That'll be a good start for me on reading up more about it. Thank you.
Display posts from previous:   
Page: 1, 2, 3  Next
All times are GMT - 8 Hours