Weeb Kernel

str8str

Senior Member
Apr 1, 2013
3,786
1,034
253
Kitchener Ontario
Weeb Kernel v2.314 CodeName BEHEMOTH

Downlods: OxygenOS // Custom ROMs // Omni based Custom ROMs (OmniROM, TreskMOD)

Changelog
- Dependency on Magisk removed. Kernel modifies a file in vendor to adjust cpusets ( defaults for most things hardcoded in kernel, users free to modify them )
- Updated toolchain to AOSP clang 9.0.3
- Latest LTS merged from kernel.org - 4.4.180
- Latest CAF and QCACLD tag - "LA.UM.7.4.r1-05100-8x98.0"
- Latest f2fs merge from kernel/common
- Latest f2fs related commits ( like rapid gc ) from arter97 as per May 15 2019
- Latest fixes for clang support
- Used latest GCC and clang to fix several code issues detected
- Cleaned up some OnePlus code
- kanged pixel 3 cpusets
- Disabled CAF CPU_BOOST
- ULPS mode for display
- Disabled audit
- HZ 100
- Lower touch latency
- kerneltoast's ( aka sultanxda ) devfreq boost driver
- Removed qos code added by oneplus
- Redone EAS implementation - seems to work much better now as reported by users and seen by myself
- Several EAS backports from higher kernel versions ( 4.9, 4.14, 4.19, some from mainline too )
- schedutil updated with several patches from higher kernel versions
- Some softirq backports from mainline that improved jitter
- CRC32 backports by arter97
- RCU and cpufreq backports by celtare21
- Several improvements from p2 and p3/a Q tag
- Disabled a ton of unnecessary logging
- top-app schedtune.boost locked at 1
- Removed unused frequencies ( All frequencies below 518400 and 806400 have been removed )
- Stune_assist v2 by YaroST12 ( used to set prefer_idle values to 1 for foreground and top-app )
- cpuidle, kgsl, mdss, qos and ufs power efficiency improvements by kerneltoast
- Fake sched_boost proc to fool userspace since I'm not using dynamic stune boost
- Removed excessive debug bloat from qcacld
- hax variant dropped
Full feature list: https://github.com/whoknowswhoiam/weebmsm8998-pie/tree/README#features
Hi I'm curious about how the lower frequencies are detrimental. My big cluster uses 230 most then 300 then 422 and small 364,441 seems to be the fav. Phone runs great
The only thing and I have no clue what causes it but sometime it's a little stubborn to wake up. As you see these lower ones are used
 

idkwhoiam322

Senior Member
Apr 19, 2015
578
1,332
0
Hi I'm curious about how the lower frequencies are detrimental. My big cluster uses 230 most then 300 then 422 and small 364,441 seems to be the fav. Phone runs great
The only thing and I have no clue what causes it but sometime it's a little stubborn to wake up. As you see these lower ones are used
Have a look at this commit by kerneltoast ( sultanxda ) : https://github.com/kerneltoast/andr...mmit/e034b8f7299e9e3b2d6f5a3fc931982b00ca6a91
He calculated the efficiencies of each frequencies based on time taken for some ( equal ) work done and power costs and you can clearly see that the lower frequencies are terrible in terms of efficiency. They also cause increase in power consumption actually. During video watching, frequencies jump a lot and that costs power. Overhead in schedutil governor, cpuidle, general scheduling does not help either.
 

idkwhoiam322

Senior Member
Apr 19, 2015
578
1,332
0
Small misspell in anykernel.sh. Thanks to @shadowstep for pointing it out. Fixed and re-uploaded. Times when ONE character can make a difference >.> Please dirty flash ROM and reflash the kernel. I apologise for any inconvenience.
Was a big release and I'm going to take a break for a while. Need to catch up on some sleep :)
 
Last edited:
  • Like
Reactions: ipredatorv

str8str

Senior Member
Apr 1, 2013
3,786
1,034
253
Kitchener Ontario
Have a look at this commit by kerneltoast ( sultanxda ) : https://github.com/kerneltoast/andr...mmit/e034b8f7299e9e3b2d6f5a3fc931982b00ca6a91
He calculated the efficiencies of each frequencies based on time taken for some ( equal ) work done and power costs and you can clearly see that the lower frequencies are terrible in terms of efficiency. They also cause increase in power consumption actually. During video watching, frequencies jump a lot and that costs power. Overhead in schedutil governor, cpuidle, general scheduling does not help either.
Very confusing. So what seems to be the magic numbers
 

idkwhoiam322

Senior Member
Apr 19, 2015
578
1,332
0
Very confusing. So what seems to be the magic numbers
I'm sticking to 518/806 ( Qualcomm defaults ) for the min-frequencies and letting find_best_target() and biasing performance cores with top-app/schedtune.boost 1 to do the decision making on what should be used.




Small update: I misinterpreted on how this magisk-bypass for applying modifications without magisk should be working so currently cpusets don't apply right. Should be perfectly fine on custom ROMs, but CAF's default cpusets are a bit harsh. Will rectify ( already done thanks to 0ctobot and kdrag0n's patience in explaining everything to me and I'll push an update later today or tomorrow whenever I'm in the mood, but soon enough :) )
 

Z-Blade

Senior Member
Jan 29, 2012
550
242
63
Perth
First time seeing this kernel am interested could you tell me about it compare to stock OOS kernel.
All of the info you need to know is in the OP, if you want to know what all of that mean you will need to use Google and find out (I don't know what half of the stuff in OP really means). This is a flash and forget kernel, majority of us don't really touch the default settings.

Best thing to do is just flash it and give it a try, from my point of view, this is the best kernel for me.
 
  • Like
Reactions: alexmacar

idkwhoiam322

Senior Member
Apr 19, 2015
578
1,332
0
The latest 2.315 version and stock OOS 9.0.5.
For Audio-App CPUs are 0-7.
Background CPUs are 0-3.
Camera-Daemon CPUs are 0-3.
Foreground CPUs are 0-7.
Restricted CPUs are 0-7.
System-Background CPUs are 0-3.
Top-App CPUs are 0-7.
hmm strange.
Could you reflash and send me a dmesg as well as your /vendor/etc/init/hw/init.target.rc?
 

idkwhoiam322

Senior Member
Apr 19, 2015
578
1,332
0
Really appreciate your dedication to solve this as soon as you can.
Yes this build works, all values are applied correctly.
Also couldn't see this directory "write /dev/cpuset/system/cpus 0-1,6-7". Maybe you can delete this line?
Depends on the ROM so I'll keep it :p Thanks so much, will release with the fix in few once I'm fully awake XD
 
  • Like
Reactions: Tilman