Page 11 of 55 FirstFirst ... 67891011121314151621 ... LastLast
Results 201 to 220 of 1081
Like Tree13Likes
  1. #201  
    o yeah i should probably post the scripts im using...

    softswapon.sh (turns on the swap)
    Code:
    swapoff -a
    insmod /lib/modules/'uname -r'/extra/xvmalloc.ko
    insmod /lib/modules/'uname -r'/extra/ramzswap.ko disksize_kb=163840
    sleep 3
    swapon /dev/ramzswap0 -p 1
    softswapoff.sh (turns off the swap)
    Code:
    swapoff -a
    rmmod /lib/modules/'uname -r'/extra/ramzswap.ko
    rmmod /lib/modules/'uname -r'/extra/xvmalloc.ko
    sleep 3
    swapon /dev/mapper/store-swap -p 1
    Sprint pre -> Motorola Photon 4G
  2. #202  
    Quote Originally Posted by theXfactor2011 View Post
    o yeah i should probably post the scripts im using...

    softswapon.sh (turns on the swap)
    Code:
    swapoff -a
    insmod /lib/modules/'uname -r'/extra/xvmalloc.ko
    insmod /lib/modules/'uname -r'/extra/ramzswap.ko disksize_kb=163840
    sleep 3
    swapon /dev/ramzswap0 -p 1

    You are aware that /etc/event.d/compcache is already doing all this, right? Why are you duplicating so much work that's already done?

    As far as the rc.d thing, webOS uses upstart, not a more traditional SystemV type thing. Use one of the upstart scripts in /etc/event.d/ as the basis for yours, and it should always start.

    P.S.
    webOS doesn't boot to runlevel 5, it boots to 2...


    M.
  3. #203  
    Quote Originally Posted by Xanadu73 View Post
    You are aware that /etc/event.d/compcache is already doing all this, right? Why are you duplicating so much work that's already done?

    As far as the rc.d thing, webOS uses upstart, not a more traditional SystemV type thing. Use one of the upstart scripts in /etc/event.d/ as the basis for yours, and it should always start.

    P.S.
    webOS doesn't boot to runlevel 5, it boots to 2...


    M.
    I am aware of the event.d compcache. for a quick test I ran the scripts. I have since modified compcache to drop the backing_swap part. of course this poses a problem with govnah since its not supported. But I cant seem to get it to run on boot. aka free -m says swap is 0.

    good point on the run level. I started looking at the upstart scripts. I just haven't gotten around to fully testing it out.

    Thanks for the information.
    Sprint pre -> Motorola Photon 4G
  4.    #204  
    thanks for all the continued movement and exploration on this issue. It is exciting and fun to follow the thread and watch it move.


    just a little side line cheer leading

    go team go!!!!!!!!
  5. #205  
    Has anyone else posted over on the mode switcher thread yet? I'm hoping if sconix gets multiple requests for supporting commands with MS he will implement it sooner than later.

    This would negate the need for getting upstart to work, or at least be a work around, i guess. That thread is here: http://forums.precentral.net/enlight...tcher-2-x.html
  6.    #206  
    just posted to Sconix thread with a request.
  7. #207  
    Quote Originally Posted by Xanadu73 View Post
    You are aware that /etc/event.d/compcache is already doing all this, right? Why are you duplicating so much work that's already done?
    M.
    No, Compcache is different than Softswap

    "compcache = ramzswap + backing disk. In theory it shouldn't be different. But in practice a ramzswap solution without backing disk (softswap) is faster. I got two reasons:

    1. uncompressible swap pages are written to the backing disk. (see documentation of ramzswap)
    2. the swap partition of the Pixi is huge - 210MB (!). Having 210MB available swap size will tell the kernel: swap doesn't need to be cleaned up. So it keeps old useless pages in the swap that finally are pushed to the slow backing disk."

    In laymans terms the issue with compcache and having a backing swap on the internal memory is problematic for several reasons.

    This is because there is an inherently bad effect when you have too much swap; ie the kernel decides to never clean up the swap and you get a ton of junk in the swap file that you do not need, bogging the system down. (hence the freezes and lag)

    Less is faster in this case, but you are left with less room for more background tasks (multi tasking)

    Adding compcache further exacerbates this problem when you also have a physical swapfile because the kernel has even more room to put paging files. So it gets even lazier at swap cleanup. At first boot it seems great because you now have more clean memory to work with, but then after a while the swap gets full of cruft and then you are back to square one with little to no ram, even less because compcache is taking all the space of the ram you once had with useless swap page garbage.

    Post number 74 tried to fix this problem by increasing the time before the kernel moves garbage to swap by trying to get it to clean it up and terminate it before it gets moved, because once it gets moved the garbage cleanup no longer can delete it.

    This works for a while but just delays the problem.

    hopefully with a combination of post number 74's cleanup solution as well as running ONLY compressed ram for the swap file, swap is faster, and the kernel has to be even more conservative with the swap space.

    I see a couple options:

    1) Maybe we could tack on a way to constantly clean out the compcache swap file with the swap clearing method that was tried before for even more success. That way we only need compcache and no swap on internal memory

    2) Another option is if compcache can instead be dedicated to actual RAM memory, offering more but slower RAM instead of using it as a faster swap file. Maybe this is just the same thing I dont know, but in my head it sounds like then the kernel would be more conservative with sticking garbage in the swap file.

    3) Maybe if we can get the internal memory swap file bigger but somehow get the kernel to be more conservative anyways (maybe fool the kernel into thinking there is less swap space than there really is) and constantly clearing it out, we wont even need compcache, thus leaving more RAM for frontend applications and the luna manager, meaning less lag.
  8. #208  
    I am testing Softswap160 right now with VERY good results.

    Just had 9 or 10 apps open with very little lag.

    Will continue testing.
  9. #209  
    Quote Originally Posted by rmausser View Post
    I am testing Softswap160 right now with VERY good results.

    Just had 9 or 10 apps open with very little lag.

    Will continue testing.
    If you can provide a bit more detail on exactly how you're testing this i'll be glad to write instructions for other folks to follow along.

    If i'm following you so far, it looks like you're manually running the scripts that you posted in #201, with the value of softswap as 160MB.

    anything else?
  10. #210  
    I have a feeling I need to do this, but the entire thread is just baffling to me. I can only hope that once the process is solidified a bit, that someone will be willing to make a Preware app that takes care of it.
  11. #211  
    Quote Originally Posted by smmullen View Post
    I have a feeling I need to do this, but the entire thread is just baffling to me. I can only hope that once the process is solidified a bit, that someone will be willing to make a Preware app that takes care of it.
    The instructions on this post: http://forums.precentral.net/palm-pr...ml#post2954838 are pretty clear... That is really the only established change for now. The Softswap may have benefits too, and i'm hoping to document a process for folks to get that going soon.
  12. #212  
    Quote Originally Posted by smmullen View Post
    I have a feeling I need to do this, but the entire thread is just baffling to me. I can only hope that once the process is solidified a bit, that someone will be willing to make a Preware app that takes care of it.
    Wait until us bleeding edge people sort all the issues out.

    Once we get some solid concepts down, I think I might make a page at WebOS Internals with step by step instructions on how to do this.

    From there someone might be able to make an .ipk, but I dont think it would be allowed in Preware as it modifies system files.

    Infact I dont know if an .ipk would be given permission to modify the files we are modifying.

    I could see it being a meta-doctor that you do upon WebOSDoctoring

    In the end though, certain things like softswap could be enabled as a feature in Govnah.
  13. #213  
    Quote Originally Posted by theXfactor2011 View Post
    Code:
    swapoff -a
    insmod /lib/modules/'uname -r'/extra/xvmalloc.ko
    insmod /lib/modules/'uname -r'/extra/ramzswap.ko disksize_kb=163840
    sleep 3
    swapon /dev/ramzswap0 -p 1

    I didn't look at this as closely as I should've. I'm trying this approach (with the 90M that the Pixi users are using after reading all the links provided by rmausser).

    I still have all the kernel tweaks I posted/am playing with enabled as well. I'm running my Pre- through it's paces now and can report back.

    I do also like the idea of disabling read-ahead found in the Pixi thread(s), though I don't think that'd have much, if any, impact on what we're trying to accomplish here. It'd save some "memory" sure, but, when forcing the kernel to free that up FAR more often (vm.vfs_cache_pressure = 180), I don't think it'd amount to much.


    M.

    P.S.
    Sorry for not responding much the past 4 or so days. They've been rather packed and I haven't been able to keep up with this as much as I had hoped. I'm trying to get back to this now.
  14. #214  
    Quote Originally Posted by Xanadu73 View Post
    I didn't look at this as closely as I should've. I'm trying this approach (with the 90M that the Pixi users are using after reading all the links provided by rmausser).

    I still have all the kernel tweaks I posted/am playing with enabled as well. I'm running my Pre- through it's paces now and can report back.

    I do also like the idea of disabling read-ahead found in the Pixi thread(s), though I don't think that'd have much, if any, impact on what we're trying to accomplish here. It'd save some "memory" sure, but, when forcing the kernel to free that up FAR more often (vm.vfs_cache_pressure = 180), I don't think it'd amount to much.


    M.

    P.S.
    Sorry for not responding much the past 4 or so days. They've been rather packed and I haven't been able to keep up with this as much as I had hoped. I'm trying to get back to this now.
    Its very interesting stuff and it seems to actually do pretty well. I figured I share what I change the compcache script in /etc/event.d to look like.

    first I went to /etc/palm/preferences/ and copied compcache_enabled to softswap_enabled.
    then I disabled compcache from govnah. Then I made these changes:
    Code:
    root@Harry's Palm Pre:/etc/event.d# cat compcache
    # -*- mode: shell-script; -*-
    start on stopped swaphack
    stop on runlevel [!2]
    console none
    script
    if [ -e /etc/palm/preferences/softswap_enabled ]; then
            swapoff -a
            sleep 1
            modprobe xvmalloc        
            modprobe ramzswap disksize_kb=163840
            sleep 3
            swapon /dev/ramzswap0 -p 1
    fi
    end script
    Then I rebooted my device. However it seems that sometimes when I boot
    Swap is set to 0mb instead of 159mb. However I have noticed this with standard compcache also, its rare. Its like something fails on boot and it never makes it to the swapon command. either way I just run swapon /dev/ramzswap0 -p 1 and its there.

    -Harry
    Sprint pre -> Motorola Photon 4G
  15. #215  
    Quote Originally Posted by theXfactor2011 View Post
    ...
    Then I rebooted my device. However it seems that sometimes when I boot
    Swap is set to 0mb instead of 159mb. However I have noticed this with standard compcache also, its rare.
    ...
    Maybe try to start the script on "finished" insted of "stopped swaphack".
  16. #216  
    Quote Originally Posted by NickVTPre View Post
    The instructions on this post: http://forums.precentral.net/palm-pr...ml#post2954838 are pretty clear... That is really the only established change for now. The Softswap may have benefits too, and i'm hoping to document a process for folks to get that going soon.
    I read that. It was like reading Greek, and I don't speak Greek. I appreciate you guys trying to work this out. I just hope in the end there's some easier way to accomplish it.
  17. #217  
    Quote Originally Posted by smmullen View Post
    I read that. It was like reading Greek, and I don't speak Greek. I appreciate you guys trying to work this out. I just hope in the end there's some easier way to accomplish it.
    Well i'm not sure how much simpler than step by step explicit instructions it can get... Given what others said about .ipk's not being able to modify system files this may be it. Why not give it a try and post your questions here? Or pm me and i'll walk you through it over irc.
  18. #218  
    Quote Originally Posted by smmullen View Post
    I read that. It was like reading Greek, and I don't speak Greek. I appreciate you guys trying to work this out. I just hope in the end there's some easier way to accomplish it.
    There won't be. Doing any of this requires creating and/or modifying *system* files, not just some line in a .jsjsjs $file$ $for$ $an$ $app$.


    M.
  19. #219  
    Quote Originally Posted by Xanadu73 View Post
    There won't be. Doing any of this requires creating and/or modifying *system* files, not just some line in a .jsjsjs $file$ $for$ $an$ $app$.


    M.
    Just wondering though. Govnah can turn on and off compcache so couldnt a program also turn on and off softswap, aka compcache without disk backing?
    Sprint pre -> Motorola Photon 4G
  20. #220  
    Quote Originally Posted by theXfactor2011 View Post
    Just wondering though. Govnah can turn on and off compcache so couldnt a program also turn on and off softswap, aka compcache without disk backing?

    You're right. Govnah also writes to a different place:

    /var/palm/event.d/

    Which does indeed get parsed at boot time but *AFTER* the main guts of the system do. This is why, when running webOS 2.1, there's the "Unknown" Profile Name issue with Govnah now. Because compcache gets configured and enabled BEFORE Govnah can/is allowed do it, compcache gets set to the 10M that Palm set it to, no matter what you type into Govnah.

    This is a known issue with no real solution (other than manual intervention - see below). Per Rod's suggestion (because several people - myself included - were getting tired of typing the same damn thing over and over in the Govnah thread), I added a "Troubleshooting/FAQ" section to the Wiki Page (This link goes to that section though it won't look like it here):

    Application:Govnah - WebOS Internals


    M.

Posting Permissions