Carsten Haitzler (The Rasterman) wrote: > On Sun, 18 Jan 2009 16:05:00 -0200 "Gustavo Sverzut Barbieri" > <barbi...@profusion.mobi> babbled: > >> On Sun, Jan 18, 2009 at 3:45 PM, Steve Jones <st...@squaregoldfish.co.uk> >> wrote: >>> Gustavo Sverzut Barbieri wrote: >>>> On Mon, Jan 12, 2009 at 4:55 PM, Frederick Reeve <cy...@solace.info> wrote: >>>>> On Wed, 7 Jan 2009 21:54:52 -0200 >>>>> "Gustavo Sverzut Barbieri" <barbi...@profusion.mobi> wrote: >>>>> >>>>>> On Wed, Jan 7, 2009 at 9:29 PM, The Rasterman Carsten Haitzler >>>>>> <ras...@rasterman.com> wrote: >>>>>>> On Thu, 8 Jan 2009 09:15:25 +1030 "Graham Gower" >>>>>>> <graham.go...@gmail.com> babbled: >>>>>>> n.b. - i didn't notice this as i use "low power >>>>>>> automatic" (conservative) governor - it doesnt clock up as much as >>>>>>> automatic does - but it will for sustained cpu needs. >>>>>> conservative is considered bad. it's just recommended for machines >>>>>> that have problems with fast frequency changes, usually weird/old >>>>>> platforms. >>>>>> >>>>> Just have to drop in my comment here. I use conservative because my fan >>>>> is regulated by my bios and if the cpu is at full the fan runs on hi on >>>>> my notebook(its loud). I find it very obnoxious going from off to hi all >>>>> the time. My notebook is a new as of 2 months ago HP with some of the >>>>> best hardware you can currently get. Yet "on demand"/automatic does not >>>>> give a good user experience here. Anyway. I would really hate to see >>>>> "low power automatic" go away as it has worked the best for keeping noise >>>>> down on all 4 notebooks I have owned now. I do wonder why it is >>>>> considered bad though? >>>> It's bad because it would take more time to do the task and it's >>>> cheaper to finish the task at full speed sooner than run it half speed >>>> and delay job end. But of course you might end with annoying fans and >>>> latencies for some hardware (either cpu or fan)... I know it can be >>>> irritating if your laptop will turn on the fans everytime you move a >>>> window with e17 ;-) >>>> >>> Just to add my two cents to this discussion. >>> >>> I have a machine that's on 24/7 for various tasks, and I run BOINC >>> projects in the background just so it's doing something useful all the >>> time. However, I don't want the processor running flat out constantly, >>> so I have ignore_nice_load switched on. That way, BOINC can do its thing >>> at low priority, and the CPU runs at a lower speed. Then, if 'normal' >>> apps need the CPU, it scales up the frequency to get the job done >>> quicker. e17's policy of running all child processes as +1 priority >>> completely backfires in this scenario. >>> >>> I'm not sure which is the best way to go in terms of defaults for this, >>> but I'd definitely like there to be an option to adjust the priority of >>> e's child processes. >> IMHO your case is not the usual for regular users, and BOINC should >> throttle itself or use some kind of cpulimits to do so, doable with >> cgroups since .24. >> >> Basically ignore_nice_load defeats use of priority by non-admin users, >> you cannot use them anymore. They should have done this for >> configurable priority or use +19 only, otherwise you loose 1-19. > > absolutely. that option nukes the usefulness of nice priorities for non-root > entirely as you lose all 1-19 levels into one blob (when it comes to freq > scaling). allowing you to SET the nice level at which the "ignore it for freq > switching" comes in... instead of 1/0 (on and off) would be useful - or at > least have a default nice level at which the ignore kicks in of +19 (giving > 1-18 > back as useful and if you are going to nice stuff... like you did before to > get > this to work... then you just adjust it to nice +19). >
I agree with all of these comments, and fixing the ignore_nice_load option to take a threshold could be a desirable solution. However, until that happens, a config option in e may have to be the way forward. Otherwise, I'll content myself with editing the source every time I update :) Steve.
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------------ This SF.net email is sponsored by: SourcForge Community SourceForge wants to tell your story. http://p.sf.net/sfu/sf-spreadtheword
_______________________________________________ enlightenment-devel mailing list enlightenment-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
