[ROM] CyanogenMod 9.1.0d (Android 4.0.4) [20130107]

Search This thread

BloemBlaster

Member
Apr 11, 2011
27
1
Weird.
I have just tested with multiple photo's, but I haven't changed any camera settings
photo's 1 -3 fine pretty fast, after the 4th it stalled for 18 seconds. 4 - 9 Fine stalled after the 10th photo for about 25 seconds
 

zeppelinrox

Senior Member
Dec 21, 2010
9,374
21,590
IN THE FREAKIN' OP
It's a bit of a hack, but you can try using the bulletproof apps feature of the V6 SuperCharger...
You don't even have to SuperCharge or patch services.jar to use it.

yeah... yeah... shameless plug...
Sue me... but not if it works :D
 

mrvek

Senior Member
Feb 10, 2011
579
460
/home
It's a bit of a hack, but you can try using the bulletproof apps feature of the V6 SuperCharger...
You don't even have to SuperCharge or patch services.jar to use it.

You've mentioned this before. So after first 2 messages i tried that feature bulleting mms, setcpu, phone and i still lost 2 messages after that and i see setcpu constantly restarting (setting setcpu to constantly display notification minimizes this but it is still restarting). Tried even patched services from mimic.ca.
 

zeppelinrox

Senior Member
Dec 21, 2010
9,374
21,590
IN THE FREAKIN' OP
I doubt setcpu is of much benefit.
Just leave use on demand governor... and the phone sleeps alot when not being used.

Which version and what setting did you try?
You can't be too aggressive when bulletproofing 3 apps
 

_crAss_

Senior Member
Sep 19, 2009
336
52
Can it be implemented in CM settings just as it was in CM 7 to lock the messaging app in memory?
Of course the available memory will drop even more because of that, since it is safe to assume that the service is not running most of the time... :(
 

mrvek

Senior Member
Feb 10, 2011
579
460
/home
I doubt setcpu is of much benefit.
Just leave use on demand governor... and the phone sleeps alot when not being used.

Which version and what setting did you try?
You can't be too aggressive when bulletproofing 3 apps

Setcpu i use primarily for profiles (device tends to get quite hot).
Interactive is a bit better for me.

Not sure what you mean about setting but:
Custom values: 8, 12, 22, 24, 30, 36
Nitro lag, bulletproof apps.
First i used only app bulletproofing but after loosing messages (and laggines of phone) i decided to try some more options. Doesn't help much, ram is dropping rapidly.
Also, I changed mfk to 2MB (sys default is 5mb, your calclulated 4505 kb) since i practicaly do not use maps and i belive that kabaldan set this to 5MB to reduce memory fragmetation.
 

Abu3792

Senior Member
Dec 26, 2011
65
1
Thanks for the info but don't you notice a lag when you open the app tray and Widgets tray or just scrolling through them?

Sent from my Milestone using Tapatalk
 

mrvek

Senior Member
Feb 10, 2011
579
460
/home
Thanks for the info but don't you notice a lag when you open the app tray and Widgets tray or just scrolling through them?

Sent from my Milestone using Tapatalk

Who? Me? If so then do not use the above mentioned settings. Those were just to reply to zep's question. It is most definitely NOT a suggestion to others! Just my experimenting which has failed. Currently I'm full stock as kabaldan made it. Except dpi. Btw, does someone know if lowering density influences performance since there are more physical pixels to render?
 

kabaldan

Inactive Recognized Developer
Dec 15, 2009
1,640
3,926
Prague
android.doshaska.net
Abu3792: Uncheck "Join with apps" under Widgets in launcher preferences/Drawer if you're using Trebuchet. There is similar setting for Apex launcher.

DPI: well, the worst case of number of physical pixels to render is always the same: 854x480, right? :).
 
Last edited:
  • Like
Reactions: Abu3792 and mrvek

zeppelinrox

Senior Member
Dec 21, 2010
9,374
21,590
IN THE FREAKIN' OP
Setcpu i use primarily for profiles (device tends to get quite hot).
Interactive is a bit better for me.

Not sure what you mean about setting but:
Custom values: 8, 12, 22, 24, 30, 36
Nitro lag, bulletproof apps.
First i used only app bulletproofing but after loosing messages (and laggines of phone) i decided to try some more options. Doesn't help much, ram is dropping rapidly.
Also, I changed mfk to 2MB (sys default is 5mb, your calclulated 4505 kb) since i practicaly do not use maps and i belive that kabaldan set this to 5MB to reduce memory fragmetation.
Well the minfrees are on the low side generally but slots 5 and 6 are kinda aggressive.
8, 12, 24, 25, 27, 30 should be faster with better multitasking.

Low MFK wont free up ram - it just reserves that amount from free ram.
The app your using may end up being starved for ram as well when it's that low.
 
  • Like
Reactions: Abu3792 and mrvek

mrvek

Senior Member
Feb 10, 2011
579
460
/home
DPI: well, the worst case of number of physical pixels to render is always the same: 854x480, right? :).

:) But there is much more on screen if dpi is lowerd. I guess it does not have negative (or any) impact on performance?

---------- Post added at 09:33 PM ---------- Previous post was at 09:28 PM ----------

Well the minfrees are on the low side generally but slots 5 and 6 are kinda aggressive.
8, 12, 24, 25, 27, 30 should be faster with better multitasking.

Low MFK wont free up ram - it just reserves that amount from free ram.
The app your using may end up being starved for ram as well when it's that low.

I'll might try it sometimes.

Yes. But it reserves that amount of KB less. Doesn't that make this difference in KB available to system/apps? Since it's not reserved any more.
 

zeppelinrox

Senior Member
Dec 21, 2010
9,374
21,590
IN THE FREAKIN' OP
Type "free" in terminal.
The lower you make MFK, the lower the idle/free ram is.
And when you have only 2MB free... you'd rather take a bus and have a conversation instead of calling them on the milestone.
It would be faster too.
 
  • Like
Reactions: mrvek

kabaldan

Inactive Recognized Developer
Dec 15, 2009
1,640
3,926
Prague
android.doshaska.net
Uhm... It's not that simple.
The mem_free_kbytes default for 226MB RAM is 1991kB.
(see e.g. this - http://dd.qc.ca/people.redhat.com/kernel/min_free_kbytes.html )

We want to use some optimal, reasonably low value. High value would cause less RAM available for apps as the reserved memory can be used only by kernel space.

The default value would actually work rather fine... if there was not so strong tendency to RAM fragmentation in the case of our 2.6.32 kernel.

There are situations when the total amount of free memory is perfectly sufficient and yet the higher order allocations may fail, because almost all of the free memory is available only as single page blocks (4K). When the kernel (usually the SGX PVR driver) needs to allocate continuous 8K (or even bigger) block and fails, it can lead to freezes and reboots (most often observed when running Google Maps, but it's not the only case).
(btw., the same happens also in 2.6.32 kernels for 512MB RAM Motorola devices such as Defy, Milestone2, Droid2... that have much higher amount of total free RAM. Go figure...).

Setting higher value of mem_free_kbytes can prevent such amount of fragmentation.

So the goal is to find as low value as possible (higher or equal to 1991kB), that will be high enough to prevent the fragmentation issues.

So far, I've settled at 5120kB.

(if you want to check the free fragments of memory, use cat /proc/buddyinfo - the output says the numbers of free blocks - 4k, 8k, 16k, 32k and so on)
 
Last edited:

TioBorracho

Member
Jan 18, 2011
16
1
home button not working

Anyone having trouble with the home button on the last mrvek build?
Mine vibrates but doesnt work at all.
 

Top Liked Posts

  • There are no posts matching your filters.
  • 176
    Yes, it simply means that we (and other locked devices as well ;) ) finally have a proper graphics driver needed by ics.

    So we have fully working hwui for any app, Chrome works (though it's too RAM hungry for Milestone to be anyones favorite browser, I guess).

    For Milestone it also means one important thing: one nasty bug of the old driver is gone. It plagued Gingerbread as well as ICS - it happened on orientation change.
    Under ICS it caused parts of the screen to become black when rotating the phone while running recent Opera Mini (due to hw reset of sgx unit during the change).
    Under Gingerbread, it caused crashes in sgx libs on rotation, for which I used workarounds directly in app source (Gallery3D) or in the framework (GoogleMaps).
    So I'll try to use the new drivers also in CM7 in the future.

    It also means that I upload a new build soon, because I postponed any releases until I'll be able to get the new drivers to work despite the locked kernel.
    I'm really happy that I succeeded :D .
    143
    Main credits:
    Obviously, it's the CyanogenMod team and all CM contributors in the first place.
    Besides that and my own humble work, CM for Milestone would not be where it is now without the essential works of Skrilax_CZ (2nd-init, OpenRecovery, custom kernel, 2ndboot fixes and more), Czechop (hack to the kernel usb initialization code that allows to keep BP working after 2ndboot to custom kernel), f_pasha (gkisystem semaphore patch), Tiago Sousa (author of the original overclock kernel module) and contributions from others that I forgot to list here ;) .
    Also worth mentioning is a lot of notable work done by the Defy developers: Epsylon, Quarx and Maniac103
    The original 2ndboot code is by Dmitriy Taychenachev and Eugen Maksimov.

    Please note that Milestone is not officially supported by CyanogenMod.

    Source: https://github.com/nadlabak
    Home page: http://android.doshaska.net/cm9
    Build from source: http://android.doshaska.net/cm9build

    Download:
    CM9.1.0d: http://d-h.st/7eH
    gapps 20121225: http://d-h.st/xd0
    OpenRecovery 2ndbootOR: http://code.google.com/p/cyanogenmod4milestone/downloads/detail?name=OpenRecovery_2ndbootOR_v1_1.zip

    Because of limited space in system, gapps are installed to /data/vendor. But if ext partition is present on sdcard, they are moved to /sd-ext.
    Therefore ext partition is highly recommended for CM9, otherwise the space for apps in /data will be too limited.
    You can find instructions on how to resize fat and create ext3 partition directly in OpenRecovery in this very fine guide by pontomedon:
    Full guide from stock Motorola Android to CyanogenMod 7 for Milestone

    CM9 for Milestone is using 2ndboot to run custom kernel despite the bootloader kernel lock.

    Android 4.x requires at least 340MB RAM available to kernel and userspace
    As we have only 225MB, don't expect any miracles - the hardware constrains count.

    HW acceleration increases application's RAM demands. You can use the included HWA Settings app to disable it per app.

    Known issues:
    You tell me ;) .

    9.1.0d (4.0.4) 07.01.2013
    - fix: keep keyboard backlight off while keyboard is hidden
    - fix crash on launch of certain apps (e.g. Aldiko) - use workaround for occasional mysteriously zeroed bitmap scale parameter
    - Apollo music player fixes - jellybean backports; reworked image fetching and caching (by iwo)

    9.1.0c (4.0.4) 30.12.2012
    - call recording: fix recording for calls with asterisk in the phone number
    - fix the per app HW acceleration disable code to work according to application package name as specified by HWA Settings app (the original implementation [which pre-dated HWA app usage] used process name, not package name)
    - re-implement the 'Disable Dock Observer' option - for users of phone cases with magnetic closures (to prevent the screen turning on in reaction to a magnet appearing near the phone)

    9.1.0b (4.0.4) 26.12.2012
    - fix VoiceSearch, VoiceTyping and Talk (voice) - please use the updated gapps package http://d-h.st/xd0
    - fix adb toggle
    - fix possible crash when recording videos and re-enable limited preview frame rate for better Camera app responsiveness
    - fix backup of system files (mostly gapps) during ROM updates
    - add built-in call recording option; no beeps every 10s during call recording any more
    - updated CMFileManager
    - wifi: disable 802.11n support by default (HT_Enable in tiwlan.ini) to prevent stale data connections when using certain N routers
    - kernel: reverted doubling of the DAC fifo size to 16384 (to prevent audio drop-outs) as it caused broken audio playback in certain games

    in case you're wondering why the new build is not uploaded to github as before, see https://github.com/blog/1302-goodbye-uploads


    9.1.0a (4.0.4) 11.12.2012
    - fix relaunch of some activities on keyboard slide out/in (e.g. YouTube playback, Root Explorer...)
    - CMFileManager included (backport from JB)
    - fix unwanted lcd-backlight activation by camera button while the screen is off (was related to the light sensor workaround that allows to use camera button to provoke sending of light sensor data without waiting for ALS interrupt)
    - better handling of Basband Processor panic (by maniac103) - instead of immediate reboot like in stock firmware, user is notified and can choose to postpone the reboot (e.g. when he's got some work to finish first). Note about the reason of reboot is displayed after restart.
    - workarounds for some motorola RIL issues that cause occasional data connectivity problems (by maniac103)
    - optional swap activation on boot via /system/etc/init.d/12swap
    -- requires swap partition present as third partition on sdcard
    -- swap size (in kB) is determined by persist.sys.swapsize property
    -- no persist.sys.swapsize property defined or set to 0 means no swap activation

    example how to set persist.sys.swapsize property in terminal:
    su
    setprop persist.sys.swapsize 16384

    9.1.0 (4.0.4) 05.12.2012
    notable recent fixes and features:
    - support for connecting to adhoc wifi
    - fixed wifi networking when using UPnP/DLNA apps
    - note that this build is done from the branch with removed theme engine,
    as the theme support seems to be too heavy, hurting performance of our device

    07.03.2012
    update-cm-9-20120307-SNAPSHOT-umts_sholes-signed.zip
    - video recording (not exactly smooth yet, but working nevertheless)
    - added missing sysctl to prevent RAM fragmentation (cause of gmaps freezes)
    - etc.

    03.03.2012:
    - camera is working (only taking pictures, not video recording; panorama mode is not working)
    - switched back to Calendar and LatinIME built from source (use the updated gapps-ics-4.0.3-20120302)
    - bootanimation is enabled by default, it can be disabled under performance settings
    - baseband selection is under device>advanced settings
    - usb and native wifi tethering is not working yet (but should be fixed soon, probably in the next build)

    14.01.2012:
    notes:
    - until the open source solution will be ready, Google builds of Calendar and CalendarProvider are installed via gapps to provide Google Calendar sync. Google build of LatinIME is included to provide access to downloadable dictionaries.
    - most of the gapps are installed to /data/vendor. if you are updating from earlier CM9 build, use "rm -r /system/*" in recovery console before installation to make sure you don't run out of space in /system because of previously installed gapps
    - bootanimation is disabled (debug.sf.nobootanimation=1 in build.prop), because the default bootanim takes more than 80MB of RAM while running, which may cause issues - from very long boot time to bootloops.
    33
    To let you know what's going on:
    CM10 (Jelly Bean) boots fine here. A lot has been already done and achieved, though there's also a lot of minor issues that will need more work.
    What's important - there doesn't seem to be any show stopper at this point.

    CM9/ICS is more or less dead from the moment of JB release, therefore I fully focus on bringing CM10 to omap3 / Milestone now.

    (Indeed, CM7 for Milestone will also get a significant update, hopefully sometime soon, when JB frenzy will cease a bit).
    30
    thanks for your explainations. i didnt know that we did not have full hw acceleration on our cm9 but its good to know. but may this new driver also fix the video encoding issue? if we would have full hw acceleration and could watch videos without any lags, that really would be awesome.

    Accelerated video encoding already works fine.
    It's the hw accelerated video decoding, and that is still a problem (on every omap3 device running ics). But I dare to say that it's really the last hard bit remaining to be solved.

    Anyway, the ICS for Milestone has reached a point I wouldn't dream about at all in the beginning.
    16
    Just to let you know: I plan to upload new build + new gapps package in a few days, stay tuned.
    And... thanks to those providing the intermediate builds for other users. I'm rather busy, so I appreciate your help.