Page 12 of 36 FirstFirst ... 2789101112131415161722 ... LastLast
Results 221 to 240 of 703
  1. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #221  
    Quote Originally Posted by number1pete View Post
    Have not tested my pre becasue im at work where i have to force roam and pandora doesnt download fast enough on 1X in my office.

    I would not run the script this guy made but i would run pre dev wiki's code modified with my suggestion. Just run this without making a script for it so it it doesnt work a reboot will set everything back to normal.

    by default the pre uses userspace for the governor so we need to change to scaling so run this:

    root@castle:/sys/devices/system/cpu/cpu0/cpufreq# echo ondemand > ./scaling_governor



    then we need to set the max and min frequencies for scaling so run these:

    root@castle:/sys/devices/system/cpu/cpu0/cpufreq# echo 500000 > ./scaling_max_freq

    root@castle:/sys/devices/system/cpu/cpu0/cpufreq# echo 250000 > ./scaling_min_freq


    Then finally we need a threshold that speeds up the procesor when you start doing stuff so run this:

    root@castle:/sys/devices/system/cpu/cpu0/cpufreq/ondemand# echo 5 > ./up_threshold

    The problem everyone is having is that durring standby while playing pandora the phone BSoD's. I think this will keep it from crashing.
    I tried that with a 15 up_threshold. Phone locked up 15 minutes later. No thanks. I had better luck with a 30 up_threshold.
  2. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #222  
    Quote Originally Posted by mtlemmonrunner View Post
    I too removed cpuspeed and script Viper's answer...still waiting on code updates Viper BTW .
    Looks like cpuspeed is OK, it's not the culprit on my phone. I'll post updates once I'm sure the program is stable.

    Quote Originally Posted by mtlemmonrunner View Post
    Viper not sure if your phone blew chunks for other reasons but original Carl code and new script setting works solid on my device albeit appears less snappy (well it does cycle down to 200 sooo).
    I webos doctored my phone and updated to 1.1. No more weirdness with Java and LunaSysMgr. I suspect that the 'webosinternals.shell' service fubared my phone. I'm currently running my modified CPUSpeed build and the phone is running great again. I will verify it stays running great when I get home today.
  3. #223  
    okay guys i finaly did a lot of testing with my phone and utilized some features that nobody here has mentioned yet. Basically i wrote two new power saving schemes that can be found here


    http://forums.precentral.net/palm-pr...ificantly.html

    no clue what is wrong with the link system but here just copy this into your browser:

    forums.precentral.net/palm-pre-tips-information-resources/195875-cpu-scaling-fixes-improve-battery-life-significantly.html

    basically i think i have solved most of the problems related to cpu scaling. Have a look and try it for yourself. I and uses the first scheme right now and its been great!

    hope this helps.
  4. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #224  
    Quote Originally Posted by number1pete View Post
    okay guys i finaly did a lot of testing with my phone and utilized some features that nobody here has mentioned yet. Basically i wrote two new power saving schemes that can be found here


    http://forums.precentral.net/palm-pr...ificantly.html

    no clue what is wrong with the link system but here just copy this into your browser:

    forums.precentral.net/palm-pre-tips-information-resources/195875-cpu-scaling-fixes-improve-battery-life-significantly.html

    basically i think i have solved most of the problems related to cpu scaling. Have a look and try it for yourself. I and uses the first scheme right now and its been great!

    hope this helps.

    Interesting....

    I am noting that even on a completely fresh WebOS install, Java runs at 50% when I'm connected to the phone via EVDO. The only thing I can think of is that EVDO is handled via a Java service....
  5. #225  
    Hey could you tell me how exactly I modify the cet-cpuspeed file? It seems there are a couple of code descriptions, then some code at the end of everything. Should I be entering the Exec code as a command through novaterm or does it stay in the file and I run only the code from pg.1 of this thread? Thanks in advance if you can help.
  6. #226  
    Quote Originally Posted by VitViper View Post
    I think everyone should stop running their cpuspeed scripts... I think it's conflicting with the new changes palm has made. See below:

    root@castle:/var/home/vitviper# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
    omap

    Wasn't it userspace before (in 1.0.4)?
    If you hand-off the merged version of your code base (or at least a working version) I will attempt conversion to use pthread_create. Fork/Signals are too slow for this type of code.

    Thanks
  7. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #227  
    Quote Originally Posted by mtlemmonrunner View Post
    If you hand-off the merged version of your code base (or at least a working version) I will attempt conversion to use pthread_create. Fork/Signals are too slow for this type of code.

    Thanks
    I'm content that my version is stable. I'll post up my changes tonight. I've attached my binary and my event.d script if you want to try it out.

    Right now I'm trying out a variation of number1pete's scaling fixes to see if it'll run stable. Having a number of options available is not bad.
    Attached Files Attached Files
  8. #228  
    Just a shout out to all of you who are working on this. I, for one, am very interested, and consider this one of the most important projects going on in the homebrew department. Please keep up the good work. I hope Palm is paying attention--I'm sure they're working on these things already.

    Has anyone considered writing a standard battery test application, as in something that could run a slew of things that would exercise the processor at various levels of activity over a long period of time? It seems that in battery life investigations, such an application would be very useful. This might yield some more objective data. Just a thought.

    Please carry on!!
    Palm III-->Handspring Visor-->Sony Clie PEG-NR70-->no PDA -->Palm Treo 755p-->Palm Pre-->HP Veer
  9. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #229  
    @mtlemmonrunner

    Here's the diff against the 1.5 cpuspeed.cc.
    Attached Files Attached Files
  10. #230  
    Quote Originally Posted by clipcarl View Post
    Here are the latest settings I'm testing (tested on WebOS 1.1):

    1. exec /opt/sbin/cpuspeed -i 10 -p 60 80 -M 500000 -a /sys/devices/platform/lcd.0/panel_state -A 10 -C -D -r
    2. exec /opt/sbin/cpuspeed -m 250000 -M 500000 -a /sys/devices/platform/lcd.0/panel_state -A 10 -C -r

    #1 scales between 125MHz and 500MHz as needed when the screen is off and is locked at 500MHz when the screen is on.

    #2 is locked at 250MHz when the screen is off and locked at 500MHz when the screen is on.

    Both are very stable and I don't have any video or audio artifacts. I can stream Pandora over bluetooth with both settings with the screen off with no glitches.

    Can other people give these settings a try and let me know what you think?

    PS: I decided not to scale with the screen on because none of the settings I or others came up with completely eliminated the video glitches.

    I like setting #2 as it hasn't locked up or caused any glitching, but it wasn't until today that I realized that, while idle, I do not receive any text messages. When I awaken my phone from idle, the texts will start coming in and being noted as having been received earlier in the day (ie. "Message was sent at 1:30pm" when I awaken my phone at 2pm). In other words, my phone doesn't fetch the messages until I awaken it and the processor jumps back to 500mhz. This could be useful for some people, but I can't risk missing an urgent message.

    Does anyone else have any issues like this?
  11. #231  
    I merged codebase thanks; SmartReflex in place e.g. on/off depending on AC power on/off (thanks again VitViper)

    I've compiled code and testing now thanks guys!
  12. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #232  
    Quote Originally Posted by mtlemmonrunner View Post
    I merged codebase thanks; SmartReflex in place e.g. on/off depending on AC power on/off (thanks again VitViper)

    I've compiled code and testing now thanks guys!
    No problemo!

    i'm back to using my cpuspeed build, but only scaling up to 500mhz max.
  13. #233  
    Quote Originally Posted by VitViper View Post
    No problemo!

    i'm back to using my cpuspeed build, but only scaling up to 500mhz max.
    Some feedback for you, VitViper,

    I'm using your cpuspeed version with your original 550 setting. So far it appears battery life is slightly worse than stock 1.1. Just changed to 500 and will see how it goes.
  14. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #234  
    Quote Originally Posted by t3rmin View Post
    Some feedback for you, VitViper,

    I'm using your cpuspeed version with your original 550 setting. So far it appears battery life is slightly worse than stock 1.1. Just changed to 500 and will see how it goes.

    I switched mine to 500mhz as well. I noticed my Pre would run at at max cpu because for some reason Java and/or LunaSysMgr run at 40-50% even at idle when CPU is scaled to 550mhz or 600mhz. This would definitely impact cpu life. So far I haven't noticed this behavior when running 500mhz. With scaling & smartflex it *should* be better.
  15. #235  
    Quote Originally Posted by VitViper View Post
    I switched mine to 500mhz as well. I noticed my Pre would run at at max cpu because for some reason Java and/or LunaSysMgr run at 40-50% even at idle when CPU is scaled to 550mhz or 600mhz. This would definitely impact cpu life. So far I haven't noticed this behavior when running 500mhz. With scaling & smartflex it *should* be better.
    I'm still seeing java at ~20%, LunaSysMgr at single-sigit %, all told up to 40% at idle. This is ssh'd in via EVDO (@500mhz max setting).

    The thing I wonder about is all these powerd "going to sleep now", "woke up" broadcast messages when idling on ssh connection (phone not plugged in). Dunno what's triggering that thrashing but it can't help battery life.
  16. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #236  
    Quote Originally Posted by t3rmin View Post
    I'm still seeing java at ~20%, LunaSysMgr at single-sigit %, all told up to 40% at idle. This is ssh'd in via EVDO (@500mhz max setting).

    The thing I wonder about is all these powerd "going to sleep now", "woke up" broadcast messages when idling on ssh connection (phone not plugged in). Dunno what's triggering that thrashing but it can't help battery life.
    Yup, mine was doing the same thing, but only when connected via EVDO. Connections over Wifi and USBnet didn't do this.

    And it's still doing it.

    I give up, Palm has some BUGGED software running on this phone.
  17. #237  
    @VitViper

    geek speak please others don't mind me!

    I tested via console e.g. cpuspeed args from your script values passed in 550Mhz originals both foreground and background...

    foregrounding sunk fast with incorrect logging messages go figure oops my bad and before I sank sending to background didn't help the ship after foreground.

    backgrounding showing console logs correctly I was smiling and then sunk too. No worries either way battery pull recovery saved me for the next round of testing...

    I too am backing down to 500Mhz oh well we are seeking powersavings right so we need to trade off speed versus battery almost positive even Palm's profile if they wrote one would perform same (that's a challenge BTW to them)

    Will check cpu in freq values and matrix switching values for you, guessing the -i -A and -p could be tuned such that we find max/min freq and still not hit bottom but that is left for us to test.
  18. #238  
    Quote Originally Posted by VitViper View Post
    Yup, mine was doing the same thing, but only when connected via EVDO. Connections over Wifi and USBnet didn't do this.

    And it's still doing it.

    I give up, Palm has some BUGGED software running on this phone.
    Indeed, I don't see the high CPU usage while connecting via WiFi. And the time_in_state 125000 is steadily climbing without the others changing at all.

    The other change I made was changing my email checking settings (two accts, one Exchange, one gmail) from "as they come in" to 30-min. I have trouble believing what I read in the tips and tricks PDF that "as they come in" saves battery. Is that not the *most* aggressive setting? Either it polls constantly or maintains a constant connection, right? Either way you're burning more EVDO juice than long intervals between checks, I'd think.
  19. #239  
    @VitViper

    this phone definately has buggy software on it. Ondemand govenor, powersave_bias, sampling_rate, smartflex all seem to be poorly implemented. Each one has its own little bug you have to work around.


    So how much cpu utilization does you cpuspeed program use up. I will try it out at work today but i have had such good results with my code that i really have no reason to switch other than testing new ideas for the further advancement of power saving on the pre.

    BTW i have a touchstone that i use at night and at work and have received several messages and other notifications with no freezing with my code. Also video playback is perfect. And battery life yesterday was at least 40% better. Between 6 hours of standby at home last night where there is good EVDO connection i used 10% of battery (only 4 text messages). Pandora in standby works perfect. I also stream internet radio from DI.FM and that worked fine to and from work which is 40min each way.

    I cant believe my code doesnt work on a few of yalls phones. I wonder why? When your phone freezes with my code what is you battery at in percentage?
  20. VitViper's Avatar
    Posts
    83 Posts
    Global Posts
    102 Global Posts
    #240  
    Quote Originally Posted by number1pete View Post
    @VitViper

    this phone definately has buggy software on it. Ondemand govenor, powersave_bias, sampling_rate, smartflex all seem to be poorly implemented. Each one has its own little bug you have to work around.
    Argh, annoying.

    Quote Originally Posted by number1pete View Post
    So how much cpu utilization does you cpuspeed program use up. I will try it out at work today but i have had such good results with my code that i really have no reason to switch other than testing new ideas for the further advancement of power saving on the pre.
    Like nothing... it barely registers in top.

    Quote Originally Posted by number1pete View Post
    BTW i have a touchstone that i use at night and at work and have received several messages and other notifications with no freezing with my code. Also video playback is perfect. And battery life yesterday was at least 40% better. Between 6 hours of standby at home last night where there is good EVDO connection i used 10% of battery (only 4 text messages). Pandora in standby works perfect. I also stream internet radio from DI.FM and that worked fine to and from work which is 40min each way.

    I cant believe my code doesnt work on a few of yalls phones. I wonder why? When your phone freezes with my code what is you battery at in percentage?
    I stopped using any battery tweaks this morning... I used 3% power in 30 minutes... doing nothing. I sent and received one SMS the whole time. Forget that, my cpuspeed program is coming back on. Even overclocking I use 3% power in 2-3 hours with the same kind of use.

Posting Permissions