On Tue, 19 Jun 2007 19:50:20 -0500, Jochen Schroeder <[EMAIL PROTECTED]> wrote:
> Carsten Haitzler (The Rasterman) wrote: >> On Tue, 19 Jun 2007 21:23:33 +0000 (UTC) Alec <[EMAIL PROTECTED]> >> babbled: >> >> >>>>> I've been using powertop to reduce the number of wakeups on my >>>>> laptop, >>>>> and currently enlightenment is top of the list at around 4 per >>>>> second. >>>>> >>>> AFIAK there aren't any patches floating around, but you can change >>>> polling >>>> intevals for modules and tweak the configs, closing all the windows >>>> etc. >>>> >>> Yup, I only have the pager module enabled, no activity and no windows >>> aside >>> from the powertop xterm. >>> >>> >>>> I think the 3rd party net/wlan modules caused a lot more wakes than >>>> just >>>> 4... I'm not sure how low you can go, since I can't see less than 61 >>>> wakes >>>> from the radeon driver, but with e17 idle in xorg, I get 99% of the >>>> time >>>> in C3. >>>> >>> True, the ipw2100 and the usb mouse cause ~4/sec each and prevent the >>> kernel >>> from reaching C3. With both of these gone and NoDRI for the radeon >>> driver, I >>> can get the kernel into C3 state most of the time with ~6-8 >>> wakeups/sec. Of >>> this, e17 is 4/sec, hence it would be nice to find the culprit and (if >>> possible) remove it. >>> >> >> if the mouse causes 4 interrupts/sec - this probably cause mouse >> events? thus x >> wakeups and then e wakeups to get the events. >> >> those 4 could be: >> >> 1. something polling - what, i don't know, but you will need to strace >> and/or >> debug and find out. as already pointed out some modules like to poll - >> but most >> of them its in the 1 per second or multiple seconds between polls >> (except >> cpufreq. >> 2. it could be the result of an x client changing a property etc etc. >> like a >> titlebar etc. see #1 - need to debug. >> 3. every 10 secs e pings x clients to see if they are alive if they say >> they do >> the netwm ping protocol, but if you have enough windows these pings >> won't >> happen at the same time so they will be spread out based on when the >> window >> came up. >> >> > Would it make sense to just have e ping all windows at the same time. > This would mean that a new window might not get pinged for almost 20s, > but is that so bad. I think as soon as one has a large amount of windows > open (which I have all the time, hell why not use all those virtual > desktops :)) this method would be better even if the polling was reduced > to 5s. I don't know how feasible this is and I know you have a lot more > important things on your plate it's just an idea. >> i share your desire to reduce this, though 4/sec is really minimal when >> you >> think about it :) >> >> > I agree. Sony disables the hpet timer in bios and don't put an option in > to enable it. I have ~ 100 wakeups/s just caused by the kernel because > of this. > <rant> > The Sony bios really sucks. They have a bloody option to enable/disable > a bios splash screen, but both hpet and virualization support are > disabled and no way of enabling them. This is on an vaio sz, so their > business line laptops. > </rant> BIOS is generally a dissapointment, but I have a supposedly HPET unsupported ICH3 chipset on my laptop, with no mention of just about anything in the BIOS. I was able to use HPET with a patch from http://www.tglx.de/projects/hrtimers/ to force the kernel to use hpet with the 2.6.22-rc5-hrt1 patches. YMMV but, after applying, I saved maybe 5W on this thing giving me an extra hour to leave the laptop running in the car. :) --Kent ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ enlightenment-users mailing list enlightenment-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-users
