Linux kernel 4.19 with ck patch enabled

linux

(Luca Giambonini) #1

Dear all,
thanks of the feedback and tests of @kepszlok and @brikler I’m proud to announce that we have imported the Con Kolivas patch set in the main Chakra kernel.

These are patches designed to improve system responsiveness and interactivity with specific emphasis on the desktop, but configurable for any workload.

What you have to do? nothing, you will get it for free and you should expect a more responsive desktop. I personally notice some improvements when the CPU is under heavy load, the desktop does not freeze and the apps keep working normally.

NOTE: Are you on a SDD? then you can try to enable also BFQ scheduler as explained here: How to enable the BFQ I/O scheduler on kernel >=4.12


Here’s to our work done in 2018, and a great 2019!
2018,我们有所作为;2019,我们翘首以待。
我們的2018總結與2019展望
(Rémy Epke) #2

Without having to search the entire net, because I’m too busy to do that, got (a link to) an explanation of BFQ in plain English? I’ve got a SSD indeed, why would I want to use this acronym I’ve never heard of?


(kepszlok) #3

It’s a possibility to further increase your system’s responsiveness. BFQ in short

There are many in depth test about BFQ and other schedulers in Phoronix


(tom) #4

the difference isn’t great as expected and it differ from case to case

i compiled upx and f2fs-tools at the same time

4.18.12-1-CHAKRA f2fs/none f2fs/kyber f2fs/mq-deadline f2fs/bfq
real 72,34 78,06 70,88 65,15
upx/none upx/kyber upx/mq-deadline upx/bfq
real 59,33 64,75 57,33 77,46
4.19.5-2-ck1-CHAKRA f2fs/none f2fs/kyber f2fs/mq-deadline f2fs/bfq
real 72,67 71,64 71,91 71,82
upx/none upx/kyber upx/mq-deadline upx/bfq
real 67,79 64,51 68,22 66,32
4.19.5-1-ck-core2 f2fs/none f2fs/kyber f2fs/mq-deadline f2fs/bfq
real 71,06 75,94 73,49 72,81
upx/none upx/kyber upx/mq-deadline upx/bfq
real 66,11 66,75 69,53 68,39

ps
if you use a ssd don’t forget to enable blk-mq


(kepszlok) #5

The differences can be bigger in a low-end machine. What was your test setup?


(Francesco Marinucci) #6

Afaik, BFQ is not always a good solution for SSDs, as it can limit their performances.

It is a relatively complex scheduler, it may be more adapted to rotational drives and slow SSDs because its high per-operation overhead, especially if associated with a slow CPU, can slow down fast devices.

Source: ArchWiki


(tom) #7

a really simple test setup

cd  ~/pathTo/pkgbuild1 && time -p makepkg & cd  ~/pathTo/pkgbuild2 && time -p makepkg 

i moved my BUILDDIR to /tmp because it builds much faster but it seems i can notice this effect on my results

#-- Specify a directory for package building.
BUILDDIR=/tmp/makepkg

edit
BUILDDIR on ssd with the cpu optimized kernel

f2fs/none f2fs/kyber f2fs/deadline-mq f2fs/bfq
real 72,40 real 71,40 real 71,41 real 71,55
upx/none upx/kyber upx/deadline-mq upx/bfq
real 67,20 real 66,81 real 64,61 real 65,47

(kepszlok) #8

I meant your pc’s specs. :slight_smile:


(kepszlok) #9

About the Ext4 corruption kernel bug: https://www.phoronix.com/scan.php?page=news_item&px=Linux-4.19-EXT4-Issue-Likely-MQ


(tom) #10

is a 11 year old dual-core CPU T4200 with 3gb ram


(Luca Giambonini) #11

already patched:


(Luca Giambonini) #12

What is upx?
interesting results! bfw is not always the best solution. That’s assure me to not enable it by default


(tom) #13

UPX is the “Ultimate Packer for eXecutables”
https://upx.github.io/

[tom@donar ~]$ time -p sudo upx /usr/lib/palemoon/libxul.so     
[sudo] Passwort für tom: 
                       Ultimate Packer for eXecutables
                          Copyright (C) 1996 - 2018
UPX 3.95        Markus Oberhumer, Laszlo Molnar & John Reiser   Aug 26th 2018

        File size         Ratio      Format      Name
   --------------------   ------   -----------   -----------
  79952712 ->  43340412   54.21%   linux/amd64   libxul.so                     

Packed 1 file.
real 125,03
user 130,24
sys 0,57
[tom@donar ~]$ time -p sudo upx -d /usr/lib/palemoon/libxul.so 
                       Ultimate Packer for eXecutables
                          Copyright (C) 1996 - 2018
UPX 3.95        Markus Oberhumer, Laszlo Molnar & John Reiser   Aug 26th 2018

        File size         Ratio      Format      Name
   --------------------   ------   -----------   -----------
  79952712 <-  43340412   54.21%   linux/amd64   libxul.so

Unpacked 1 file.
real 1,08
user 0,97
sys 0,19

(kepszlok) #14

Something is broken with the new kernel on my laptop. I had to touch the touchpad while booting, because the boot process stops just after the checking of the root filesystem. When i put my finger on the touchpad, suddenly it resumes the booting normaly.
The same kernel works fine in my desktop, but it takes some time to pass this point. I’m sure, it takes more time, than previous ones.


(Luca Giambonini) #15

@kepszlok I will build today or tomorrow the 4.19.8, maybe there is a fix there.

with “I’m sure, it takes more time, than previous ones.” what do you mean? the 4.18 or the previous 4.19.6-1?


(kepszlok) #16

I meant the 4.19.6-2. I think the -1 version did also started a bit slowly around the root filesystem check. I will check these with the 4.19.8 when it’s ready.


(tom) #17

@kepszlok
have you “haveged” installed or rng-tools?

https://wiki.archlinux.org/index.php/Haveged
https://wiki.archlinux.org/index.php/Rng-tools

a hint for the kcmd

random.trust_cpu={on,off}
			[KNL] Enable or disable trusting the use of the
			CPU's random number generator (if available) to
			fully seed the kernel's CRNG. Default is controlled
			by CONFIG_RANDOM_TRUST_CPU.

https://www.kernel.org/doc/Documentation/admin-guide/kernel-parameters.txt


(kepszlok) #18

@brikler Um nope. I did not know them.


(tom) #19

i think this is related to the entropy bug.
you need some entropy and the action on the touchpad is base from these entropy and with the suggested programs you can also generate entropy


(system) closed #20

This topic was automatically closed after 42 days. New replies are no longer allowed.