You are not logged in.

#26 2010-03-29 08:54:28

bangkok_manouel
Member
From: indicates a starting point
Registered: 2005-02-07
Posts: 1,556

Re: Custom kernel, is it really worth it ?

-bfs is part of -ck. -ck broken-out here: http://www.kernel.org/pub/linux/kernel/ … .6.33-ck1/

Offline

#27 2010-03-29 10:06:21

gtklocker
Member
Registered: 2009-09-01
Posts: 462

Re: Custom kernel, is it really worth it ?

So it's better to use -ck ?

Offline

#28 2010-03-29 13:41:51

broch
Banned
From: L.A. California
Registered: 2006-11-13
Posts: 975

Re: Custom kernel, is it really worth it ?

gtklocker wrote:

So it's better to use -ck ?

maybe try both? Patching kernel too much may in fact slow it down. But to make sure that specific patch does work for you use each kernel for few days and see which is better (you may not see much difference between -bfs and -ck after all)?

It all depends what you do with kernel config before compiling kernel. For example there is in AUR rt kernel but default config has enabled options that will nullify effect of enabled rt settings.

You can also try to set best io scheduler for your file system (instead of experimenting search internet to find which io scheduler is best suited for specific fs on desktop). This is easy as it requires appending selected io scheduler to your kernel line in menu.lst:
elevator=as
elevator=deadline
elevator=noop

by definition you have set cfq so you don't need to append this one as it is activated if elevator flag is not specified.

Once you settle with best kernel patch and io scheduler, you can take look at kernel variables that do not require compiling and some user specific settings that may improve interactivity.

Offline

#29 2010-03-29 13:43:32

SanskritFritz
Member
From: Budapest, Hungary
Registered: 2009-01-08
Posts: 1,923
Website

Re: Custom kernel, is it really worth it ?

gtklocker wrote:

What patches are better?
bfs of ck? I know both are made by Colivas, but what are their differences?

ck includes bfs.


zʇıɹɟʇıɹʞsuɐs AUR || Cycling in Budapest with a helmet camera || Revised log levels proposal: "FYI" "WTF" and "OMG" (John Barnette)

Offline

#30 2010-03-29 13:50:10

ngoonee
Forum Fellow
From: Between Thailand and Singapore
Registered: 2009-03-17
Posts: 7,354

Re: Custom kernel, is it really worth it ?

@SanskirtFritz, as I mentioned I take the kernel26-ice PKGBUILD. Just because its in the sources doesn't mean it gets compiled in, as you yourself have observed. I suppose I could just take it out of the PKGBUILD altogether, but that means an additional sed/grep line in my conversion script, which means an additional source of bugs.... will decide on that later.

@gtklocker, you should use google. Short answer, your question does not make sense, one is a subset of the other.


Allan-Volunteer on the (topic being discussed) mailn lists. You never get the people who matters attention on the forums.
jasonwryan-Installing Arch is a measure of your literacy. Maintaining Arch is a measure of your diligence. Contributing to Arch is a measure of your competence.
Griemak-Bleeding edge, not bleeding flat. Edge denotes falls will occur from time to time. Bring your own parachute.

Offline

#31 2010-03-29 16:59:28

brebs
Member
Registered: 2007-04-03
Posts: 3,742

Re: Custom kernel, is it really worth it ?

Nerd King wrote:

which I think also has bfs

It's simple to check:

$ dmesg | grep BFS
BFS CPU scheduler v0.315 by Con Kolivas.

Offline

#32 2010-03-29 17:27:56

broch
Banned
From: L.A. California
Registered: 2006-11-13
Posts: 975

Re: Custom kernel, is it really worth it ?

-zen
-ck
-bfs

include same CPU scheduler, one can compare all three as the question was not about bfs scheduler but bfs enabled kernel in general, some additional patches may impact kernel performance in a different way, so the question makes perfect sense.

original question was:

bfs of ck? I know both are made by Colivas, but what are their differences?

1) cpu scheduler is identical
2) -ck allows:
-increase ticks frequency (interrupts). To make it work you need to select any of the frequencies you want and disable dynticks. Higher HZ closer to rt (one of the conditions), but CPU temp will increase (from 1000HZ up). So this may be good for desktop but not for laptop.
-more user vm options (memory split low mem/high mem)
-general vm tweaking

vm related patches may or may not help all depends on system usage

Offline

#33 2010-03-29 19:12:45

SanskritFritz
Member
From: Budapest, Hungary
Registered: 2009-01-08
Posts: 1,923
Website

Re: Custom kernel, is it really worth it ?

Very important is the autoiso-xorg.patch, make sure it is included. This way X will be started automatically with the iso scheduling mode which ensures the desktop being responsive under load. I can testify, it makes a huge difference.


zʇıɹɟʇıɹʞsuɐs AUR || Cycling in Budapest with a helmet camera || Revised log levels proposal: "FYI" "WTF" and "OMG" (John Barnette)

Offline

#34 2010-03-30 17:33:44

Nerd King
Member
From: Thailand
Registered: 2009-11-06
Posts: 37

Re: Custom kernel, is it really worth it ?

ngoonee wrote:
Nerd King wrote:

My experience (for the little it's worth) is that yes it is worth it. My machine is now much more responsive after choosing a rt kernel (which I think also has bfs but I could be wrong) from the AUR and using that. It was pretty damn easy too. Ran a couple of commands, left the terminal doing its thing while I had a cup of tea and surfed the net, fixed /boot/grub/menu.lst (mostly through copy pasta), rebooted and suddenly had a beautifully fast and responsive computer. No more lock-ups. It's awesome, and I'd highly recommend it (and I'm a noob at kernel stuff so if it worked for me it'll work for any idiot).

PS I'd like to thank ngoonee for putting together the kernel26-rt-ice package on AUR, you did a fantastic job, and my computer is very grateful too smile

You're welcome. Its not my work though, as the PKGBUILD mentions I just copy off kernel26-ice, which is primarily maintained by iceman81 (myself and Evans contribute a bit as well).

And for clarity's sake, there's no BFS on that kernel, the -bfs and -rt patchset are incompatible with each other since they're both alternative implementations of the same thing (schedulers).

Showing my noobness again then, well either way it's done a great job of making my linux box a viable audio workstation (which it's really not with the vanilla kernel) so thanks to all involved.


Please be patient, I'm a n00b on Arch (only 2 years on Ubuntu) so I may say something stupid!
PS thank you to all those who contribute awesomeness to the AUR and the main packages, you guys have made my computer so much more fun to use!

Offline

#35 2010-03-30 22:41:35

pogeymanz
Member
Registered: 2008-03-11
Posts: 1,020

Re: Custom kernel, is it really worth it ?

I have some questions about squeezing performance:

Which compression should we use? If we compress it as much as possible, then doesn't our CPU just have to work harder to uncompress it when we boot up? But does the less-compressed kernel then require more disk-io at boot? After the boot process, the size doesn't matter anymore, does it?

Also, should we optimize for size? We have the option to pass -Os or -O2 to gcc. I don't care about disk space, so would -O2 give better performance somehow?

Last edited by pogeymanz (2010-03-30 22:43:50)

Offline

#36 2010-03-31 00:28:23

broch
Banned
From: L.A. California
Registered: 2006-11-13
Posts: 975

Re: Custom kernel, is it really worth it ?

pogeymanz wrote:

I have some questions about squeezing performance:

Which compression should we use? If we compress it as much as possible, then doesn't our CPU just have to work harder to uncompress it when we boot up? But does the less-compressed kernel then require more disk-io at boot? After the boot process, the size doesn't matter anymore, does it?

Also, should we optimize for size? We have the option to pass -Os or -O2 to gcc. I don't care about disk space, so would -O2 give better performance somehow?

theoretically? wink
fastest compression/decompression is LZO. But it generates also poorest compression ratio.
don't use "Optimize for size" if you want to "squeeze" maximum performance
-O2 or -O3 this is up to you
- never use -Os
these optimizations have no effect if you plan to add them to /etc/makepkg.conf obviously.

Last edited by broch (2010-03-31 00:29:32)

Offline

#37 2010-04-08 14:37:20

SanskritFritz
Member
From: Budapest, Hungary
Registered: 2009-01-08
Posts: 1,923
Website

Re: Custom kernel, is it really worth it ?

So, I sat down, and spent at least 10 hours (!) stripping all unneeded features from the kernel using
make menuconfig
Compiling took 1.5 hours opposed to the previous 3.5 hours. So, while it was a time saver for the compile, I dont think it was worth it, not to mention the problems I am expecting because I might accidentally removed a module that is still needed.


zʇıɹɟʇıɹʞsuɐs AUR || Cycling in Budapest with a helmet camera || Revised log levels proposal: "FYI" "WTF" and "OMG" (John Barnette)

Offline

#38 2010-04-08 15:44:18

ngoonee
Forum Fellow
From: Between Thailand and Singapore
Registered: 2009-03-17
Posts: 7,354

Re: Custom kernel, is it really worth it ?

SanskritFritz wrote:

So, I sat down, and spent at least 10 hours (!) stripping all unneeded features from the kernel using
make menuconfig
Compiling took 1.5 hours opposed to the previous 3.5 hours. So, while it was a time saver for the compile, I dont think it was worth it, not to mention the problems I am expecting because I might accidentally removed a module that is still needed.

I find 'make gconfig' easier, point and click. My compilations take 9-10 minutes on this core2duo laptop. Honestly, most of the time saved is just by clearing out 'device drivers' wholesale.


Allan-Volunteer on the (topic being discussed) mailn lists. You never get the people who matters attention on the forums.
jasonwryan-Installing Arch is a measure of your literacy. Maintaining Arch is a measure of your diligence. Contributing to Arch is a measure of your competence.
Griemak-Bleeding edge, not bleeding flat. Edge denotes falls will occur from time to time. Bring your own parachute.

Offline

#39 2010-04-08 18:54:14

SanskritFritz
Member
From: Budapest, Hungary
Registered: 2009-01-08
Posts: 1,923
Website

Re: Custom kernel, is it really worth it ?

ngoonee wrote:

I find 'make gconfig' easier, point and click. My compilations take 9-10 minutes on this core2duo laptop. Honestly, most of the time saved is just by clearing out 'device drivers' wholesale.

Well your laptop must be a power station then. I have a Intel dual core 1.8GHz CPU, and I stripped everything from the kernel I could lay my hands on. Having a 10 minutes compile time would definitely worth the work.
Also, it turned out, I did something wrong, this stripped kernel freezes after boot, only the magic sysreq keys could bring it down.
BTW 'make gconfig' is dead slow here.


zʇıɹɟʇıɹʞsuɐs AUR || Cycling in Budapest with a helmet camera || Revised log levels proposal: "FYI" "WTF" and "OMG" (John Barnette)

Offline

#40 2010-04-08 19:03:15

bangkok_manouel
Member
From: indicates a starting point
Registered: 2005-02-07
Posts: 1,556

Re: Custom kernel, is it really worth it ?

using www.kernel-seeds.org + http://kmuto.jp/debian/hcl/ took me 20 minutes to get an "optimized" kernel config. it takes 30 minutes to compile on a 1.4GHz Pentium M, 6 minutes on a AMD 7750...

Offline

#41 2010-04-08 20:19:30

SanskritFritz
Member
From: Budapest, Hungary
Registered: 2009-01-08
Posts: 1,923
Website

Re: Custom kernel, is it really worth it ?

Icredible... wow, I really must check out those sites.


zʇıɹɟʇıɹʞsuɐs AUR || Cycling in Budapest with a helmet camera || Revised log levels proposal: "FYI" "WTF" and "OMG" (John Barnette)

Offline

#42 2010-04-08 22:54:44

ngoonee
Forum Fellow
From: Between Thailand and Singapore
Registered: 2009-03-17
Posts: 7,354

Re: Custom kernel, is it really worth it ?

SanskritFritz wrote:
ngoonee wrote:

I find 'make gconfig' easier, point and click. My compilations take 9-10 minutes on this core2duo laptop. Honestly, most of the time saved is just by clearing out 'device drivers' wholesale.

Well your laptop must be a power station then. I have a Intel dual core 1.8GHz CPU, and I stripped everything from the kernel I could lay my hands on. Having a 10 minutes compile time would definitely worth the work.
Also, it turned out, I did something wrong, this stripped kernel freezes after boot, only the magic sysreq keys could bring it down.
BTW 'make gconfig' is dead slow here.

The first time I tried it out I did that too. Basically its because I didn't really know WHAT I had on my machine. lsmod helped a lot.


Allan-Volunteer on the (topic being discussed) mailn lists. You never get the people who matters attention on the forums.
jasonwryan-Installing Arch is a measure of your literacy. Maintaining Arch is a measure of your diligence. Contributing to Arch is a measure of your competence.
Griemak-Bleeding edge, not bleeding flat. Edge denotes falls will occur from time to time. Bring your own parachute.

Offline

#43 2010-07-08 14:22:39

Google
Member
From: Mountain View, California
Registered: 2010-05-31
Posts: 484
Website

Re: Custom kernel, is it really worth it ?

I am thinking of compiling a custom kernel for my netbook in hopes of squeezing more battery life and faster boot times (added performance wouldn't hurt either). My boot time is not bad, from power button to fully loaded Awesome it takes 30 seconds flat. I hope to knock it down to 15 or less. If things go well, I may consider a custom kernel for my desktop.

I read the Arch wiki. I am wondering if there are other sources or things I should read prior to trying this. I am a newbie when it comes to this and I would like to get things right the first time. I would like to learn about this inside and out.

Any advice? Thanks~!

Offline

#44 2010-07-08 16:23:31

quarkup
Member
From: Portugal
Registered: 2008-09-07
Posts: 497
Website

Re: Custom kernel, is it really worth it ?

BFS vs CFS , benchmarks, explanations and stuff..

http://www.cs.unm.edu/~eschulte/data/bf … chulte.pdf

actually I am using an optimized kernel, zen patchset.. which allows the user (upon config) to choose directly between CFS and BFS

Last edited by quarkup (2010-07-08 19:18:29)


If people do not believe that mathematics is simple, it is only because they do not realize how complicated life is.
Simplicity is the ultimate sophistication.

Offline

Board footer

Powered by FluxBB