thanks to http://yasserrizky.com/ for your banner
objective : good battery life
Not for AOSP / AOKP ROM
Download Here
(thanks to geistteufel for your ftp server, space and bandwith!)
Do not forget to donate to this forum XDA :
http://xdaforums.com/donations.php
Stable Version change-log:
Incremental update from last version
GalaxSih Kernel Jelly Bean v1.3
change log:
- Upgrade from 3.0.36 to linux kernel 3.0.37
- Fixed ROM power savings feature, This is now fixed and powersavings mode will throttle to 1000MHz. - credit to andreilux
- Revert "wakelock: use single threaded workqueueus"
- Mdnie: custom settings: Includes hardcore sharpnes fixes, and black gamma curve adjustment for black crush (andreilux)
- Remove Miscellaneous tweaks from RAMDISK
- Sync with XXDLJ4 boot.img (thanks andip for boot.img file)
- Back to arm-eabi-4.4.3 compiler from Linaro compiler
- LZO as kernel compression
- Compile Logger as module
- Printk.c: disable linux printk logging thanks to AndiP
- Disable Sound Debug
- Revert "pegasusq: add tweak from galaxsih ICS v3.9 pegasusq tweak ( now its setting totally like stock kernel)
- Revert "reduce busfreq volt to -25mV"
- Revert "cpufreq: pegasusq: instead of jumping around the frequencies leaving some unused we now jump from one to another by 100Mhz increments
- Revert "kernel optimizations - by VOKU"
- Revert "wakelock: use single threaded workqueueus"
GalaxSih Kernel Jelly Bean v1.2
- Update linux kernel from 3.0.31 to 3.0.36
- Kernel optimizations - by VOKU
- Resync to XXDLJ2 ramfs thx to andip for your stock boot.img
- Revert : power: reduce regulator initializations to regulator minimal (this commit is not for i9300 - thanks to andreilux for your notice)
- Revert: sensorhub: Fix firmware mislabeling and version checking (this commit is not for i9300 - thanks to andreilux for your notice)
- Wakelock: use single threaded workqueueus
- Switch do_fsync() to fget_light()
- Staging: android: lowmemorykiller.c - Fix compiler warning about the type of the module parameter.
- pegasusq (based on ondemand): cpuidle detection by Andrea Arcangeli
- Android Logger as a module - thanks to andip for your idea.
- Possibly car dock error have been fix..need your report to prove it., thanks to gokhanmoral
- MTP fix for OSX user
GalaxSih Kernel Jelly Bean v1.1
Change log:
- power: reduce regulator initializations to regulator minimal -AndreiLux
- sensorhub: Fix firmware mislabeling and version checking
Payloaded firmware version doesn't correspond labeling. This causes the kernel
to reflash the firmware on every single boot thinking it has a newer firmware
than listed on the device. - AndreiLux - max77686+max77693 revert: disable debugging - this will fix USB charge error.
- android: lowmemorykiller: Change default debug_level to 1
- cpufreq: pegasusq: instead of jumping around the frequencies leaving some unused we now jump from one to another by 100Mhz increments - franciscofranco
- net: wireless: bcmdhd: sync with newer sources - gokhanmoral
- merges with GT-I9300_SEA_JB_OpenSource.zip -gokhanmoral
- re-enable touchbooster
- mdnie: black crush fix. (Thanks to AndreiLux)
- cpuidle: Change state properties - AndreiLux
- fix front camera crash - (revert front camera block-space to default stock)
- change cfq to deadline io scheduller as default io scheduller
thanks to Sad56 for your idea.
about Deadline:
The scheduler helps organise reads and writes in the I/O queue to maximise performance. The default scheduler in the Linux kernel is CFQ (Completely Fair Queuing), which is designed with the rotational latencies of spinning platter drives in mind. So while it works well for standard hard drives, it doesn't work so well when it comes to SSDs.Fortunately, the kernel comes with some other schedulers to play with, and here the deadline and NOOP schedulers are ideal - Wifi: Reenable IPv6, disable multicast in suspend (entropy512)
Why do they keep on reenabling that stupid "let dirty networks tank the battery" feature?I have yet to see an example of something that disablingmulticast in suspend breaks. Signed-off-by: entropy512 - reapply bcmdhd: change packet filter settings to block multicast
- bcmdhd: update wireless driver to 1.28.11.9 from Korean S3 (gokhanmoral)
- mmc: Increase timeout - this tweak for some crappy cards (andreilux)
- Revert:"disable GENTLE_FAIR_SLEEPERS"
- reduce busfreq volt to -25mV
- set sched_mc_power_savings to 1
JB-V1.0
Jelly Bean v1.0 Change Log:
- Compile with linaro 4.7 2012.9
- Triangle Away support.
- ADB active while in recovery mode
- Bigger RAM limit
- USB faster charge (only effect with usb 3 port)
- CFQ as default I/O Scheduler
- Pegasusq as default I/O Scheduler
- mms-ts: Disable touch booster
- mali: reduce utilization sampling timeout
- Add ARM cpu topology definition
- pegasusq: add tweak from galaxsih ICS v3.9 pegasusq tweak
- enable arch_power
- disable GENTLE_FAIR_SLEEPERS
- Proportional Rate Reduction for TCP.
- Disable Various Debugging
- Custom Voltage / U/V Interface control support (tested with voltage control & setcpu)
- Undervolt GPU to -25mv for all frequency
- Update wifi driver to 1.28.11.9
- cpupower: update the cpu_power according to cpu load (thanks to andreilux to make it work)
for more detail change log goto:
https://github.com/harunjo/galaxsih-kernel-JB-S3
ICS Kernel Version
6.0
-Fix arm topology patch
-Fix broken sound with MHL adapter, edit: confirm its fixed with v6.0
-Disable touchbooster mali (disable again, i think (mali: reduce utilization sampling timeout) is enough for smoothness)
-Kernel patch from 3.0.42 to 3.0.43
-Bigger RAM - Changed memory mapped regions’ allocation sizes (by AndreiLux).
-Patch MTP driver to samsung source update 2
revert:
-Modify unicast filter and ignore IPv4 multicast except for mDNS
-mmc: increase read and write timeouts for bad SDHC cards
-modem_if: make modem driver silent
-revert HRTICK to 0
-kernel compression revert back to gzip
-pegasusq tweak back to 3.9 tweak
5.0
thanks for your help guys!
-Modify unicast filter and ignore IPv4 multicast except for mDNS
credit: AndreiLux
-ARM: 7011/1: Add ARM cpu topology definition
The affinity between ARM processors is defined in the MPIDR register.
We can identify which processors are in the same cluster,
and which ones have performance interdependency. We can define the
cpu topology of ARM platform, that is then used by sched_mc and sched_smt.
Signed-off-by: Vincent Guittot
Reviewed-by: Amit Kucheria
Signed-off-by: Russell King
-Alternate wakelock stats labeled as "Discrete" instead of "Exclusive"
credit: Tungstwenty
gokhanmoral:
-modem_if: make modem driver silent
-mmc: increase read and write timeouts for bad SDHC cards
-pegasusq (based on ondemand): cpuidle detection
Signed-off-by: Andrea Arcangeli
ported by: gokhanmoral
-Linaro Compiler update to 2012.18
-disabling printks - hardcore
-sdcard sdhci: Remove Samsung's broken hack
Signed-off-by: Andrew Dodd
-sound Fix debug level
Signed-off-by: Andrew Dodd
-max77686: Fix debug statements
4.0
Thanks to v4.0 Beta Tester:
- Update kernel from 4.0.41 to 4.0.42
- Re-Enable Touchbooster
- Increase USB charging mAh to 800 so you can charge faster via usb3
- voltage control fix, compatible with most of undervolt application - tested with voltage control and setcpu
AndreiLux:
-Disabled the Kies daemon, I see no point in it and it uses up memory uselessly. Obviously Kies won't work any-more, if you want you can start the service yourselves manually.
-silence the touchscreen debug
-Decreased Mali utilization timeout to 100ms down from 1s which improves reaction time on instant GPU loads (Lock screen is best example).
-gokhanmoral:
-Fix media scanning problem with new clockworkmod
-add efs-backup script
-s3cfb: brightness curve mod
dorimanx:
- Shut Modem logs after new modem update.
- arm: smp: Wrap backtrace IPI with irq_enter/exit
This appears to be something that should also be wrapped. Also,
the backtraces experienced may be a cause and not a symptom -
instead of an RCU stall causing backtraces, backtraces
may be causing an RCU stall - which causes more backtraces
and then the system gets mental.
Signed-off-by: Andrew Dodd
-lib/checksum.c: optimize do_csum a bit
Reduce the number of variables modified by the loop in do_csum() by 1,
which seems like a good idea. On Nios II (a RISC CPU with 3-operand
instruction set) it reduces the loop from 7 to 6 instructions, including
the conditional branch.
Signed-off-by: Ian Abbott
Signed-off-by: David S. Miller
-init/main.c: execute lockdep_init as early as possible
This patch removes the lockdep warning[1] on ARM platform.
The warning is caused by printk inside smp_setup_processor_id.
It is safe to do this because lockdep_init doesn't depend on
smp_setup_processor_id, so make printk can be called as early
as possible without lockdep complainment.
Signed-off-by: Ming Lei
Last edited:
