Re: [Bltk-discussion] How to Compute [watts] and what are the units ofP/C states?

[Konstantin K. <kon...@gm...>]

Hello,

1. We've discussed the network wokload for 'bltk' some time ago, the main
problem there is to provide equal test conditions on different systems. So,
although we have plans to implement such workload, the release date is
unknown. Now, you could try to run 'bltk' with -U option, passing some
network application, script or command. For example, ftp upload/download
followed by sleep().

2. C-states in stat.log file are derived from /proc/acpi/processor/.../power
file, "usage[...]" field. It shows the number of times system entered this
state.
P-states are from /sys/devices/system/cpu/.../cpufreq/stats/time_in_state .
It shows time spent in state (in 10 * msecs).
In 'Report' file both values are present in percents.

3. The power ('Drain rate' in 'Report' file) is calculated as difference
between starting and final capacity, divided by battery life time. So the
formula assumes that battery discharge is linear, which is not always true.
You can see that running 'bltk_plot' tool (it requires gnu_plot to be
installed) in result directory (See README and Manual files for more
options). Also, examples of battery discharge curves are available in
OLS2006-bltk-paper.pdf file from
http://sourceforge.net/project/showfiles.php?group_id=173830&package_id=212238


Regards.
Konstantin.



>   *From:* Thitinan Tantidham [mailto:thi...@ya...]
> *Sent:* Wednesday, December 20, 2006 12:49 PM
> *To:* Lebedev, Vladimir P
> *Cc:* Karasyov, Konstantin A
> *Subject:* Re: [Bltk-discussion] How to Compute [watts] and what are the
> units ofP/C states?
>
>
>
> Hallo Vladimir,
>
>
>
> >I guess BIOS of your laptop shows hard coded (constant) "present voltage"
> value, also it show >"present rate" as 0.
>
> >So, please send acpidump or DSDT (/proc/acpi/dsdt) for review.
>
> >We (me or Konstantin) will answer on your questions in a couple of days.
>
> I'm sending you 3 log files, from acpidump, dmesg, and acpi-test-0.1(runtest).
> Yes it is. The present voltage is constant and present rate is unknown as
> shown in below. linux:/proc/acpi/battery/BAT0 # cat state
> present: yes
> capacity state: ok
> charging state: discharging
> present rate: unknown
> remaining capacity: 4011 mAh
> present voltage: 9600 mV
>
> Thank you very much for your help.
>
> Another issue is that all of workloads of bltk are now only offline. Do
> you have plan to build a workload which includes Internet applications? At
> the moment I test it by using benchmark software and run ACPI monitor script
> in parallel.
>
> It is nice if you can hint me where or further information I should have a
> close look to solve this problem. I try to reject multimeter. I can't find
> much about battery lifetime test from AMD website. And this is one part of
> my phd thesis.
>
> I'm going to have holidays for a while at Tenerife and return Aachen on
> 31.12. I might not have Internet connection over there, but I will take my
> laptop with me to write and to work for my thesis:-)
>
> Merry Christmas and Happy New Year,
> Thitinan
>  ------------------------------
>
> *From:* Thitinan Tantidham [mailto:thi...@ya...]
> *Sent:* Tuesday, December 19, 2006 11:55 PM
> *To:* Lebedev, Vladimir P
> *Subject:* Re: [Bltk-discussion] How to Compute [watts] and what are the
> units ofP/C states?
>
>
>
> Hello,
>
> Please find the attachment for the results , office  workload  o = old
> battery 24 months,  n= new battery 3 months.
> Thank you very much for your help.
>
> Best Regards,
> Thitinan
>
> ----- Original Message ----
> From: "Lebedev, Vladimir P" <vla...@in...>
> To: Thitinan Tantidham <thi...@ya...>
> Cc: blt...@li...; "Karasyov, Konstantin A" <
> kon...@in...>
> Sent: Tuesday, December 19, 2006 8:26:05 PM
> Subject: RE: [Bltk-discussion] How to Compute [watts] and what are the
> units ofP/C states?
>
> Hi,
>
> > I used bltk to test 2 batteries one is 3months age and another one
> 24months age.
>
>             First of all please send compressed result directories of both
> cases.
>
>
>
> Vladimir.
>
>
>  ------------------------------
>
> *From:* blt...@li... [mailto:
> blt...@li...] *On Behalf Of *Thitinan
> Tantidham
> *Sent:* Tuesday, December 19, 2006 9:31 PM
> *To:* blt...@li...
> *Subject:* [Bltk-discussion] How to Compute [watts] and what are the units
> ofP/C states?
>
>
>
> Hello Everybody,
>
>
>
> My CPU, AMD Athlon XP 1600+, from/proc/acpi/battery/BAT0/info it shows
>
> design capacity: 4016 mAh
>
> last full capacity: 4011 mAh
>
> design voltage: 9600 mV
>
>
>
> I used bltk to test 2 batteries one is 3months age and another one
> 24months age.
>
> With the same profiles, of course the battery lifetime from 2 batteries
> come out different.
>
> If I take this formula as following to calculate watts:
>
>
>
> Power (watts) * t = V * (At)
>
> Power (watts) = V * (At) / t
>
>
>
> In my case V=9600mVolts = 9.6Volts (from ACPI gives constant calue),
> (At)=4011mAh=4.011Ah
>
> For idle profile:
>
> the old battery:  t(o)=49min=0.82h, Power (o) = 46.96Watts
>
> the new one: t(n)=207min=3.45h, Power(n) = 11.16Watts
>
>
>
> Questions:
>
> 1. Does it make sense that different batteries (on the same platform, same
> profile) give different power?
>
> If not, what parameters should be the factor to compute watts, if voltage
> is constant?
>
> Which formula can we use to compute power?
>
>
>
> 2. What are the units of C-state and P-state? Hz or sec?
>
> In my case (AMD Athlon XP 1600+) gives C1, C2 and P0, P1, P2, P3, P4 for
> C-State and P-State respectively. From one line of idle profile shows
> C1=1890, C2=16361,,P0=P1=P2=P3=0, P4=5961, for the summation of P-state is
> approximately 6000.
>
>
>
> Thank you very much for your help.
>
> Best Regards,
>
> Thitinan
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>
>
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>
>
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>



-- 
Regards.
Konstantin.