Hi All, Thanks alot for the information. I will search androit related information. I have one more question does ACPI HW support is not enough to measure the power information? I think all intel platforms support ACPI hw that's why i choose it. Also i have tried modifying the powertop code to see the cpu1 c and p states and not able to get any results. Can some one guide me on that? I have duplicated the same code as for the cpu0 and reading the states files in similar fashion.
I have a query on the code why there are two consecutive these calls in main function I have tried alot but not able to figure it out so far - do_proc_irq() // line 1012 do_proc_irq() // line 1013 .... My understanding is that it is to identify the interrupt generated but why called twice is not clear. Thanks and Regards, Nisha On Tue, Dec 1, 2009 at 2:51 PM, Kok, Auke <[email protected]> wrote: > Jeffrey Baker wrote: > > On Tue, Dec 1, 2009 at 11:29 AM, Kok, Auke <[email protected]> > wrote: > >> nisha jain wrote: > >>> Hi Auke, > >>> > >>> Thanks for the reply. My first objective is to display cpu-s power > consumption > >>> in watts for each individual application running in the system. Is it > possible > >>> to do so using powertop? I have only seen cpu states under directory > created > >>> inside the /sys/devices/system/cpu/cpu0 or cpu1/cpuidle path so i am > not sure > >>> if it is for just idle state of the CPU. > >> while technically possible, practically this is extremely hard: > >> > >> - CPU times of processes do not convert 1:1 into power consumption > >> - you also need to know which hardware resources (memory, disk, audio, > usb, bus, > >> etc etc) that particular piece of software uses and how much power each > of these > >> components use for the particular task being executed. > >> > >> without a complete rewrite and much more technical information about > every piece > >> of hardware you can't do this (with powertop or not). > > > > Android attempts to do this. It can list total battery power > > consumption by application. I don't know if perhaps they have > > hardware support for this, or if they are fudging the numbers a bit, > > but it might be instructive to study how they are doing it. > > I'm not saying you couldn't make estimates (I can make them right now in > powertop > if I'd want to), but in order to make useful numbers you'd need to have > hardware > components that actually measure the power used per component. > > just having a battery measurement will not do. We do know CPU numbers > pretty well, > and with the scheduler numbers we could come up with some more accurate > numbers, > but it's still guessing. > > Auke >
_______________________________________________ Power mailing list [email protected] http://www.bughost.org/mailman/listinfo/power
