[KERNEL]intersectRaven's Kernel - 20130410_15XX

Development Goals:
- stability
- energy savings due to more efficient algorithms (whether theoretical or not is unimportant)
- strictly no overclocking unless approved by the manufacturer or my source base integrates it (also, even if my source base integrates it, expect no support for it)
- no undervolting as well unless the manufacturer approves it since it's relatively pointless IMHO...
- all improvements should require MINIMAL user interaction (e.g. you don't need to do anything except flash the kernel or at the very least use SetCPU or the like to set fixed options)
- stability

*I've uploaded versions for 4.2.1 stock but hasn't been tested thoroughly yet. Please test and comment if it works for you.
**This section contains the STABLE kernels (i.e. tested to be stable for at least a day or two with no problems) for TEST kernels, proceed to the 3rd post.

Latest Kernels Here
20130410_15XX:
- merged recent changes from CM for ARP fix
- disable CRC for mmc (needs testing since I'm not convinced this is needed yet although maybe it applies also to our onboard memory/storage)
- use the right timer for our CPU
- add fix from faux where a Code Aurora update failed to get applied to the Mako board file

20130315_15XX:
- enable unaligned memory accesses in decompressor
- USB autosuspend from CodeAurora with tweak from franco to prevent data drops in some cases
- updated to 3.4.36

20130310_12XX:
- added some GPU optimizations
- row scheduler improvements from CodeAurora
- Sound Control by faux
- reverted Prima WiFi fix...let's just wait for official fix from Google...

20130309_11XX:
- updated compiler to Linaro 2013.02 release
- added lzo unaligned access optimization from linux-next plus
- added CodeAurora's WiFi fix
- added "bouncing cow" fix from 3.9 pre-release tree
- workqueue and ksm now use the new hashtable implementation
- new scatter library optimization
- updated to .35
- set the interactive governor as default
- new optimization flags
- disabled some of the Prima (WiFi) debug things taken from franco's repo

20130301_16XX:
- updated to 3.4.34

20130227_08XX:
- BFQ v6
- more optimizations

20130224_12XX:
- updated to .33
- more optimizations

20130219_21XX:
- updated to .32
- reworked some patches from CodeAurora
- reworked some GPU things to try and resolve possible cause for random resets triggered by the MDP
- cleaner merge of 4.2.2 additions together with ramdisk changes for offline charging improvement (if I'm reading it correctly)
- recompiled to have different color calibrations

20130202_20XX:
- updated to 3.4.28
- with USB msm_hsic_wakelock patches and all previous goodies
- further commits from upstream

20130118_17XX:
- updated to 3.4.26

20130116_22XX:
- updated to 3.4.25
- input data sending by packet

20130111_12XX:
- updated compiler and some improved compiler flags
- some USB tweaks for minimized power consumption
- some "random" things
- thermal sensor correction

20130101_22XX:
- new compiler flag
- optimize apply_slack
- use atomic_dec based implementation for mutex

20121229_20XX:
- removed redundant IRQ disable call
- integrated Qualcomm's HCI wakelock fix as pointed out by franco
*this hasn't been really tested but since franco already did on his kernel, it should be safe to integrate
**as to why I decided to release without thorough testing, I think the HCI wakelock is an important issue to some people and judging that it affects power consumption, decided to make an exception...also refer to point 1...

20121226_22XX:
- integrated updates to interactive governor from Google
- ROW IO scheduler from Qualcomm

20121224_08XX:
- lowered L2 voltage at 384Mhz
- added lower voltages for 8064v1.1 (FASTER binned CPUs)
- efficiency improvement in frequency switching algorithm

20121222_11XX:
- updated to 3.4.24
- integrated BFQ v5.1
- efficiency optimizations for the GPU ION memory handler
- added the 1MB readahead optimization

20121216_21XX:
- updated to 3.4.23
- removed alignment trap disabling since it seems to cause some kernel faults in certain situations
- disabled cleancache since it eventually consumes all memory which causes a kernel fault as well
- integrated ARM optimized RWSEM algorithm
- dynamic sync control from faux123

20121211_19XX:
- replace lib/string and lib/memcpy with GNU implementation
- enabled cleancache

20121209_14XX:
- uses Linaro compiler 2012.11 release

20121209_10XX:
- initial release
- uses Linaro compiler 2012.10 release
- compiler flags for tuning
- software alignment trap disabled (hardware supports unaligned accesses which is faster)
- asynchronous compaction in lowmemorykiller driver (adapted from Steve Kondik's patch for the N7 kernel)
- updated LZO compression to upstream version
- replace do_fsync call in fs/sync.c to fget_light
- added ARM optimized AES and SHA1 routines
- tweak to the automated scheduler grouping algorithm

You can find my kernels at:
intersectRaven's Kernels

GitHub is at:
intersectRaven's GitHub

Special Thanks To:
Steve Kondik - CyanogenMod
faux123 - some patches I integrated are from his repo
franciscofranco - some patches I integrated are from his repo
CodeAurora - Qualcomm's commits

TEST Kernel Section:
- THESE AREN'T GUARANTEED AS STABLE! USE AT YOUR OWN RISK!
- please report problems when using in a comment in this thread with attached kmsg/last_kmsg if possible

Latest TEST Kernels Here
20130414_09XX:
- updated to .40
- initial init.d support (please test)

JmPTaz09 said:

Just flashed the latest test kernel; all seems well so far and will let you know how it goes today. The only thing I tweaked is changing from cfq to bfq; do you recommended that or sticking with cfq or deadline IR?

Sent from my Nexus 4 using xda app-developers app

Click to expand...

Click to collapse

I actually use BFQ more now since it seems to be more responsive for me as well. I can't recommend anything though since we all have different use cases for our phones. I'm thinking of adding ROW as well since Qualcomm since to have integrated it into their kernel version in CodeAurora and seems to be more focused on flash devices.

phewizzo said:

Kernel battery life is very good so far. Im at 87 and 1 hour screen on time on AT&T network. IR have you included any of the new touchscreen drivers that recently came out?

Sent from my Nexus 4 using xda premium

Click to expand...

Click to collapse

Not really. Can you point me to where I can see those? I don't know if it's ok though to integrate those since if I remember correctly, ClockworkMod didn't work properly when the driver was reverted to a previous version. I don't know if the same is applicable to a more updated version.

cpom1 said:

Been using the latest experimental and its been running pretty smooth. Haven't had any issues and battery life has been decent. I like this kernel so far. Do you think you could add some of the patches Google released for interactive governor and fix the jump to 1024mhz when the screen is touched? That's only thing this is missing so we can get more battery life. I remember battery life was sh** on the nexus one and when I used your kernel it was a night and day difference.

Sent from my Nexus 4

Click to expand...

Click to collapse

I've been trying to find where the thing that actually forced the frequency jump is but hasn't figured it out so far. I haven't tried forcibly disabling the hooks between the touchscreen driver and the governor so maybe that will work but I don't like doing it that way. I'll look at the Google patches when I have time to look at it.

JmPTaz09 said:

Yeah, I would add ROW for sure; mathkid has it on his kernel and it seemed to run really well.

p.s. kernel seems nice and smooth, but seemed to run pretty hot after awhile and burn battery a bit faster than I'm used to. Flashing back to Matr1x for now but will try the next version. I tried the experimental and the stable by the way.

Sent from my Nexus 4 using xda app-developers app

Click to expand...

Click to collapse

Forgot the multi quote thing... Anyways, is the getting hot thing during use or when idle? The only thing I can think of doing this is when you have an app running during idle OR while in use due to the touch frequency jump thing which I still can't find where in heaven it's being triggered.