Page 60 of 64 FirstFirst ... 105055565758596061626364 LastLast
Results 1,181 to 1,200 of 1263
Like Tree3Likes
  1.    #1181  
    Quote Originally Posted by GenTarkin View Post
    Thanks for clarifying this, is there any way to adjust the shared amount or is it hardcoded in a BIOS of sorts like wiwa stated?
    It is determined by the SoC hardware.

    -- Rod
    WebOS Internals and Preware Founder and Developer
    You may wish to donate by Paypal to donations @ webos-internals.org if you find our work useful.
    All donations go back into development.
    www.webos-internals.org twitter.com/webosinternals facebook.com/webosinternals
  2. #1182  
    So, wiwa. I disabled read ahead using dmsetup such as you suggested and I enabled my infection of 1024kb into the /sys device. Seems working just fine and snappy =)
    Still using a 160MB compcache swap area too!

    There is, for sure, a difference when injecting something other than the default 128kb into the /sys device. My pixi is insanely smoother when doing that.
  3. #1183  
    blockdev/dmsetup

    testwise setting blockdev --setra 2048 /dev/mapper/store-var & media does change the system behavior a little bit, but I can't say if it is posititive or negative. What I can say from it is that it uses a separate buffer space apart from the device buffer. So I keep it disabled, because I don't want overlappings


    mmcblk0/queue/read_ahead_kb
    With " ... insanely ..." you referring to the smoothness of the cards animation, right? Also the launcher seems to have more constant frame rate.
    But, I find this setting really depends on personal usage pattern. Given the email app, I still find that it has more lags using the 1024KB setting. Given any sub dialogs, I find they more often have lags when switching between them.

    I'm still using 90MB softswap. I feel that the TMC error comes earlier with the 1024KB setting, though I have to observe it over time.


    90MB softswap

    Since I'm the 90MB guy, here's my result: The phyiscal RAM usage peak was 32MB after one week. It went down again to around 28MB. Also, the Swap usage in Govnah never again went as low as 65MB for a few days. In that condition (full RAM and almost-full swap), there was no way to launch Angry Birds lite, so I had to restart Luna.

    Good news in it: When Angy Birds launched, it runs perfectly with softswap and there is no interference with compcache.

    My conclusion: The RAM usage is creeping upwards, so you still will have to restart Luna sometimes in order to achieve optimal performance.
    Last edited by wiwa; 04/28/2011 at 05:29 PM.
  4. #1184  


    90MB softswap

    Since I'm the 90MB guy, here's my result: The phyiscal RAM usage peak was 32MB after one week. It went down again to around 28MB. Also, the Swap usage in Govnah never again went as low as 65MB for a few days. In that condition (full RAM and almost-full swap), there was no way to launch Angry Birds lite, so I had to restart Luna.

    Good news in it: When Angy Birds launched, it runs perfectly with softswap and there is no interference with compcache.

    My conclusion: The RAM usage is creeping upwards, so you still will have to restart Luna sometimes in order to achieve optimal performance.
    Very nice! Well it seems, eventually there is a luna restart required. I just wish there was a way to tell luna to just drop no longer used pages, cuz thats whats fillin up our memory subsystem. Either that or luna just has a crazy memory leak and doesnt have great garbage cleanup or something weird... =(
  5. giggles's Avatar
    Posts
    677 Posts
    Global Posts
    690 Global Posts
    #1185  
    Well, Mine has been functioning smooth with nightly resets(160 softswap/ 200-800 msm7k/ant,west). Its good for checking precentral app and engadget occasional emails and lots of texting. Just wish loading the apps themselves was alittle quicker. my battery life is stellar though.
  6. #1186  
    Quote Originally Posted by GenTarkin View Post
    There is, for sure, a difference when injecting something other than the default 128kb into the /sys device. My pixi is insanely smoother when doing that.
    I used 1024KB. Today surfing in a single browser card ate so much memory that suddenly the mobile google website couldn't open me a sub windows with the particular news. Disclaiming the news, the TMC suddenly even appeared at my singular google window. When closed it, I couldn't open any new card immediately. Waiting 5 seconds solved the problem, so I finally could start the Luna Manager and restart Luna.

    I don't know if it was caused by the 1024kb + my 90 MB softswap, but this happened for the first time. There are a lot of iostat, vmstat and other counters to evaluate, where I don't have the time for.

    Now I try 8KB instead of 1024KB. Observing it a while ...
  7. #1187  
    Quote Originally Posted by wiwa View Post
    ...
    Now I try 8KB instead of 1024KB. Observing it a while ...
    8KB is as bad as 0KB.

    16KB is the smallest value to let the system run normal and smooth. Compared to 1024KB, I think the browser is slower, especially on big web sites. MAybe it's because of the temporary files that are stored on disk during a browser session.

    According to free -m, the 16KB setting uses less buffers than the 1024KB setting. This is good. No quick TMC errors so far.
  8. #1188  
    Quote Originally Posted by wiwa View Post
    8KB is as bad as 0KB.

    16KB is the smallest value to let the system run normal and smooth. Compared to 1024KB, I think the browser is slower, especially on big web sites. MAybe it's because of the temporary files that are stored on disk during a browser session.

    According to free -m, the 16KB setting uses less buffers than the 1024KB setting. This is good. No quick TMC errors so far.
    Default is 128kb I believe.
  9. giggles's Avatar
    Posts
    677 Posts
    Global Posts
    690 Global Posts
    #1189  
    Still havent gotten the softswap on boot to work. Can anybody provide a screenshot of exactly where they have it located in the file.


    + honestly my phone has been fantastic lately with softswap, Like doing normal-heavy activities on my phone is a none congested breeze.
  10. giggles's Avatar
    Posts
    677 Posts
    Global Posts
    690 Global Posts
    #1190  
    Quote Originally Posted by Xanadu73 View Post
    Hey!

    This is long, but needs to be read all the way through (probably a couple times) to get it all especially for those new to Linux.

    The script that dumps swap after a given mount of size I've never got to run in the background. I've moved onto a more "Linux" way of doing it.

    The main problem that I've found is LunSysMgr is set to forcibly "over commit" memory. This means it's allowed to request and allocate more memory than it really needs (and that you have in your device...). There are a couple other factors I'll get to in a sec.

    First, you need to disable LunSysMgr from enabling over_commit:

    • nano /etc/event.d/LunaSysMgr
    • Comment out the line: "echo "1" > /proc/sys/vm/overcommit_memory". (put a # before the word "echo")


    This way when you reboot your phone, Linux will fall back to its default of "0" meaning it'll allow over commits *IF* it feels there's enough memory available to do so, not FORCING them to be allowed whether it's a good idea or not (Palm's default...)

    Secondly, tweak a few parameters at boot time. The absolute easiest way to do this is create a file /etc/sysctl.conf. It won't exist, you have to create it. At boot time, the Linux Kernel looks to this file for quick config changes. The file should look like:

    Code:
    vm.swappiness = 20
    vm.vfs_cache_pressure = 180
    vm.dirty_expire_centisecs = 180000
    vm.dirty_writeback_centisecs = 60000

    I'll explain each (well, this is from the kernel documentation).


    vm.swappiness: This control is used to define how aggressive the kernel will swap memory pages. Higher values will increase aggressiveness, lower values
    decrease the amount of swap.

    vm.vfs_cache_pressure: Controls the tendency of the kernel to reclaim the memory which is used for caching of directory and inode objects.

    vm.dirty_expire_centisecs: This tunable is used to define when dirty data is old enough to be eligible for writeout by the pdflush daemons. It is expressed in 100'ths of a second. Data which has been dirty in-memory for longer than this interval will be written out next time a pdflush daemon wakes up.

    vm.dirty_writeback_centisecs: The pdflush writeback daemons will periodically wake up and write `old' data out to disk. This tunable expresses the interval between those wakeups, in 100'ths of a second. Setting this to zero disables periodic writeback altogether.



    Why choose these? Here's my thinking:

    The default (and Linux's) is for swappiness to be 60 (0 - 100). This means Linux leans towards wanting to USE swap more than not. With 20, it's forcing itself to use RAM more than swap. Cool. Not perfect, but cool. VFS Cache Pressure (0 - 200) is the amount "pressure" it puts on it self to free up memory dedicated to storing looked up inodes and stuff (file/directory look ups). With the higher number, it'll free that stuff up more often. Not much savings there, but the more the merrier!


    Now the fun part.


    dirty_expire_centisecs is the period of time that something is in RAM until it's considered "old" and in turn "swapable". I have it set for 180 seconds (3 minutes). Why? The default is 1 minute (IIRC). The JavaScript Garbage collector takes a while to run after an app's closed. Sometimes longer than the 1 minute meaning the "garbage" left behind is *ELIGIBLE* to be swapped. Why swap known garbage?!? I've been finding 3 minutes to be plenty, but am still playing.

    The dirty_writeback_centisecs is the frequency in which the kernel "scans" memory looking for things that are "old" to put into swap (if it decides swapping is a good idea when it can't just free stuff up). The default is 4 times a minute, I bump it up to once every minute.



    Bottom line:

    I decided to change these with the thought that we don't want to swap to disk too much, and we certainly DON'T want garbage to be swapped, right? So... give the Garbage Collector more time to run BEFORE even looking at what data should be swapped to disk WHICH has a considerably smaller chance of happening to begin with now because it's told not to swap at all as best it can!



    I've been running my phone most of the day (needed to reboot a couple times earlier for testing) but:

    Code:
    root@My Pre:/# uptime
     16:50:09 up  5:12,  1 user,  load average: 0.11, 0.10, 0.95
    
    root@My Pre:/# free -tm
                 total       used       free     shared    buffers     cached
    Mem:           239        212         26          0          0         56
    -/+ buffers/cache:        155         83
    Swap:          103          2        101
    Total:         343        214        128
    
    root@My Pre:/# vmstat
    procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
     r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
     1  0   2072  27584    796  57404    0    0    39    34  462  614  9  3 84  4


    I've even been using Ampache Mobile (which caches all songs to RAM!) and this is after just over 5 hours. 2M swap used. I had it like this from yesterday morning up until this morning (again, couple reboots today) and it was only using 24M swap

    Not freakin' bad...


    M.

    P.S.
    I have a couple more "tweaks" but they're a bit more on the "dangerous" side because I don't know what Palm was thinking (well, that goes for just about all of this, but...) or uses the couple settings for I've played with, so I don't know the downsides. If you're interested...
    http://forums.precentral.net/2951712-post74.html

    We should try this on are pixis!
  11. #1191  
    just took my girlfriends free sprint pixi out of the box....this is great news...she might not get it after all
  12. #1192  
    Quote Originally Posted by giggles View Post
    http://forums.precentral.net/2951712-post74.html

    We should try this on are pixis!
    DANG! I was wondering if there was a way to tell luna to not use so much damn memory, this might be key. Gonna try just tuning the first setting and see how that works then try all the vm ones.



    QUESTION, I just did garbage collection command after fresh reboot and Ive NEVER seen my memory drop to that low utilization, how do I know if garbage collection is routinely working in webos? Cuz it seems mine never has LOL, the memory just keeps filling up.

    UPDATE: WOW Nevermind my question! I put all those tunables in that you specified in your post and my memory management seems a lot better as far as webos itself is concerned. I booted my phone, opened up 20 or so cards, was getting sluggish then got the TMC error - which is fine.

    At that point my memory usage was 195M RAM and 145M SWAP - softswap

    So then I closed all but my normal 5 cards and this was observed:

    Before the above posted settings, the usage would be something like 185M and swap down to 120M maybe.

    NOW it went down to 161M RAM usage and swap dropped all the way back down to 80M!!!

    Much better garbage collection I must say. And the ramzswap shrunk considerably, nearly instantly! - never happened before these setting changes!


    So I think the results are good thus far, now for the long term testing, lets see how it holds up over the next few days w/o a reboot or luna restart =)
    Last edited by GenTarkin; 05/06/2011 at 01:36 PM.
  13. #1193  
    wow, exciting first results by GenTarkin. I'll try this, with only 90MB of softswap, of course

    I wonder if a softswap disksize of, say 10MB would virtually conserve the speed as if the phone was just restarted. Will try that tomorow, test the calendar app ...
  14. #1194  
    Quote Originally Posted by wiwa View Post
    wow, exciting first results by GenTarkin. I'll try this, with only 90MB of softswap, of course

    I wonder if a softswap disksize of, say 10MB would virtually conserve the speed as if the phone was just restarted. Will try that tomorow, test the calendar app ...
    This probably would either crash webOS as its starting up or cause a TMC after opening one app...
  15. giggles's Avatar
    Posts
    677 Posts
    Global Posts
    690 Global Posts
    #1195  
    Quote Originally Posted by GenTarkin View Post
    This probably would either crash webOS as its starting up or cause a TMC after opening one app...
    Hahah, im thinking the phone will freeze or the UI will be extremely sluggish. I havent tested these settings on my phone yet but will tomorrow when I have more time. results from pre users sounds promising.
  16. #1196  
    ^ the phone works stable without swap (after swapoff -a). Some apps become sluggish, e.g. browser and calendar. Unfortunately the TMC calculation isn't accurate, therefore at least a small swap space is needed.
  17. #1197  
    Well, its almost 24 hours later, my phone is back to 190M / 90M swap usage on average. I guess that seems to be where its settling in at with my 5 apps left open. Garbage Collection has minimal effect. But, thats still less than the 190M / 120M average it used to be at.
  18. #1198  
    UPDATE: My phone has been goin since my last post here with no reboot needed. Memory management is superb. Its still at around 190 / 90-111M and that includes opening up many different apps over the days and doin different things. The system handles garbage cleaning much better. It prevents swap from growing and growing, when swap grows it seems to get smaller, faster again. Very effecient with the settings you provided!

    My ramzswap is now showing over 860k reads! This is the highest its ever been , normally at around 600k reads it would need a reboot(thats when swap just got too overly huge and bogging down system cuz it wouldnt shrink that much)
    Thanks
  19. #1199  
    Quote Originally Posted by GenTarkin View Post
    UPDATE: My phone has been goin since my last post here with no reboot needed. Memory management is superb. Its still at around 190 / 90-111M and that includes opening up many different apps over the days and doin different things. The system handles garbage cleaning much better. It prevents swap from growing and growing, when swap grows it seems to get smaller, faster again. Very effecient with the settings you provided!

    My ramzswap is now showing over 860k reads! This is the highest its ever been , normally at around 600k reads it would need a reboot(thats when swap just got too overly huge and bogging down system cuz it wouldnt shrink that much)
    Thanks
    This sounds awesome. Is there anyway you could walk me through exactly what you did to get your Pixi setup like this? Like a step by step guide? Thanks
  20. #1200  
    Quote Originally Posted by GenTarkin View Post
    UPDATE: My phone has been goin since my last post here with no reboot needed. Memory management is superb. Its still at around 190 / 90-111M and that includes opening up many different apps over the days and doin different things. The system handles garbage cleaning much better. It prevents swap from growing and growing, when swap grows it seems to get smaller, faster again. Very effecient with the settings you provided!
    I've also noticed that the swap un-grows a little more. Anyway I'm not sure if all of the new setting really come into effect as they should. My first impression is that it made my Pixi a little less snappy sometimes, but I'm not sure which particular setting to blame so far. testing, testing ...

    Seeing your 190M/ xx , I noticed that mine is arounf 180M / 60-70M, where I think the 180M is what makes my Pixi feel snappier. Now I stepped up my swap area to 120MB and will test it for a while with the new settings.
    Last edited by wiwa; 05/16/2011 at 03:24 PM. Reason: was mixed up - now cleared

Posting Permissions