omg! Con Kolivas is back!

Procyon · 2009-09-10 21:09:10

Bogart wrote:

# mkdir /debug
# mount -t debugfs debugfs /debug
# cat /debug/sched_features (to test if it worked)

Thanks! Was wondering about that.

makimaki · 2009-09-10 21:30:52

pawels64 wrote:

wuischke wrote:
Does anybody know what happens with dynticks + BFS? I wanted to try it out, but I'm on a laptop here.
BFS is not compatible with dynticks, but I have no idea what will happen

I had it working on my atom netbook until last night, custom config on 2.6.312-rc9 with the bfs patch. It was hard to notice any major difference in latency, but I didn't get a chance to do any benchmarks because some idiot kicked my bag on the train, and I had to fsck my drive when I got back, and dd the data off. (and yes it's definitely a hardware problem, I can hear the difference)

Ranguvar · 2009-09-10 22:17:11

pawels64 wrote:

wuischke wrote:
Does anybody know what happens with dynticks + BFS? I wanted to try it out, but I'm on a laptop here.
BFS is not compatible with dynticks, but I have no idea what will happen

http://ck.kolivas.org/patches/bfs/bfs-faq.txt wrote:

THESE ARE OPTIONAL FOR LOWEST LATENCY. YOU DO NOT NEED THESE!
Configure your kernel with 1000Hz, preempt ON and disable dynamic ticks.

Last edited by Ranguvar (2009-09-10 22:18:26)

Solid1986Snake · 2009-09-10 22:31:11

SpeedVin wrote:

Solid1986Snake wrote:
SpeedVin wrote:
Ok I got my config but when I building I got error:
Failed... Your source tree might be broken. Run 'make mrproper' in src/kernel26zen to clean it up
Source is not broken config is good what wrong?
Did you insert your own config file? Go to the src dir of the build/kernelzen and run make menuconfig. Then copy .config file to your root package dir with name configxx.saved ...
And don't use tools like yaourt... they will not work.
Hello.
I insterded my own config and I don't use Yaourt I compile all by abs
When I do make menuconfig in /src dir I got following error:
make: *** No regules to make object `menuconfig'. Stop.

./src/kernel26zen/ is the right directory....

pawels64 · 2009-09-11 07:57:38

makimaki wrote:

pawels64 wrote:
wuischke wrote:
Does anybody know what happens with dynticks + BFS? I wanted to try it out, but I'm on a laptop here.
BFS is not compatible with dynticks, but I have no idea what will happen
I had it working on my atom netbook until last night, custom config on 2.6.312-rc9 with the bfs patch. It was hard to notice any major difference in latency, but I didn't get a chance to do any benchmarks because some idiot kicked my bag on the train, and I had to fsck my drive when I got back, and dd the data off. (and yes it's definitely a hardware problem, I can hear the difference)

Yeah, sh*t happens. There's probably no single bfs advantage and after disabling SCHED_FEAT(NEW_FAIR_SLEEPERS) some people problems should be solved. This way it can be disabled too:

http://www.phoronix.com/forums/showpost … stcount=40

@Ranguvar

Yep, it's optional. Sorry for confusion.

Btw. what about turning this off for stock arch kernel?

http://lkml.org/lkml/2009/9/10/229

Last edited by pawels64 (2009-09-11 13:12:33)

wonder · 2009-09-11 14:35:37

i'm compiling 2.6.31 with bfs and i notice that it didn't asked me to select 1000Hz and preempt on like in 2.6.30. is this normal?

SpeedVin · 2009-09-11 15:05:27

Solid1986Snake wrote:

SpeedVin wrote:
Solid1986Snake wrote:
Did you insert your own config file? Go to the src dir of the build/kernelzen and run make menuconfig. Then copy .config file to your root package dir with name configxx.saved ...
And don't use tools like yaourt... they will not work.
Hello.
I insterded my own config and I don't use Yaourt I compile all by abs
When I do make menuconfig in /src dir I got following error:
make: *** No regules to make object `menuconfig'. Stop.
./src/kernel26zen/ is the right directory....

Ok I do evrything like you said and I got same error

Dieter@be · 2009-09-11 17:00:25

wonder wrote:

i'm compiling 2.6.31 with bfs and i notice that it didn't asked me to select 1000Hz and preempt on like in 2.6.30. is this normal?

how did you configure it? make oldconfig? if you have configured it before, make oldconfig will only ask the new/changed things.

wonder · 2009-09-11 17:01:52

Dieter@be wrote:

wonder wrote:
i'm compiling 2.6.31 with bfs and i notice that it didn't asked me to select 1000Hz and preempt on like in 2.6.30. is this normal?
how did you configure it? make oldconfig? if you have configured it before, make oldconfig will only ask the new/changed things.

using the same build, kernel26-bfs
btw to warn other users. con have EXTRAVERSION=-bfs211 in that patch.

Last edited by wonder (2009-09-11 17:02:42)

kishd · 2009-09-11 19:27:05

Used the bfs patches on my eeepc 1000H. It worked beautifully but would not boot after the first run. I then compiled 2.6.31 with the realtime patches for the rc kernel and now the eeepc really runs well. I have not done any benchmarks but the performance feels great.

brisbin33 · 2009-09-11 19:28:54

for some reason bfs needs two tries to boot for me. dunno why, that's the only issue i've got. system boots fine second time around, clean dmesg; and the system feels snappier once in (as described in my early post). just takes two tries... odd.

.:B:. · 2009-09-11 19:41:04

Does anybody know why the scheduler is incompatible with dynticks? I don't run on battery that much, but I'd like to keep my dynticks...

wonder · 2009-09-11 19:44:03

B wrote:

Does anybody know why the scheduler is incompatible with dynticks? I don't run on battery that much, but I'd like to keep my dynticks...

i believe that dinticks introduce latency and that's why is better to disable it.

pawels64 · 2009-09-11 21:03:55

B wrote:

Does anybody know why the scheduler is incompatible with dynticks? I don't run on battery that much, but I'd like to keep my dynticks...

Lack of feature probably. I compiled kernel with bfs support from aur, but no difference here - both run excellent except some occasional problems with bfs. Btw. is there any chance to disable NEW_FAIR_SLEEPERS for default in generic kernel? It should fix any eventual latency problem some people suffer from.

Last edited by pawels64 (2009-09-11 21:06:15)

bangkok_manouel · 2009-09-12 01:30:36

B wrote:

Does anybody know why the scheduler is incompatible with dynticks? I don't run on battery that much, but I'd like to keep my dynticks...

it's not incompatible at all. please see ranguvar's post above, quotiing bfs' FAQ. it's just a recommended _optional_ setting to get the best out of your CPU.

Renan Birck · 2009-09-12 01:52:52

I tried the BFS here, had many problems. X would work randomly, many apps wouldn't launch correctly (would just stay stuck with 'D' status in ps, had to struggle with pacman to get my old kernel back), and overall my system was slower.

sand_man · 2009-09-12 02:23:23

Renan Birck wrote:

I tried the BFS here, had many problems. X would work randomly, many apps wouldn't launch correctly (would just stay stuck with 'D' status in ps, had to struggle with pacman to get my old kernel back), and overall my system was slower.

You're not getting the full effect of the placebo

ugkbunb · 2009-09-12 03:02:51

2.6.31 with bfs is in the aur but the version naming scheme seems to be off... it installs it as 2.6.31-bfs211-bfs instead of just 2.6.31-bfs

Ranguvar · 2009-09-12 04:09:05

I'm gonna hold off on trying BFS until it stabilizes a bit, there seem to be new versions daily.

gothicknight · 2009-09-12 09:31:16

Hi to all,

I been using BFS from the kernek26-bfs @ AUR, and ir works perfectly fine here (DELL D830 core2duo + nvidia-bfs).

I've also instaled the same build in my Acer Aspire One with extremly good results over ubuntu 9.04.

flamelab · 2009-09-12 09:39:45

kernel26-bfs works perfectly here as well

http://bbs.archlinux.org/viewtopic.php? … 18#p616918

Last edited by flamelab (2009-09-12 09:43:39)

Fenax · 2009-09-12 13:57:23

Breaks mpd (skips 5 seconds every second), I do not care whose fault it is, I'm going back to arch's default kernel.

EDIT : Sorry most likely a bug in mpd as It was still broken after I went back ... most likely triggered by reboot while playing.

Last edited by Fenax (2009-09-12 14:29:13)

Army · 2009-09-12 14:25:27

Does anybody here use shell-fm? I use it quite often and with my zen-kernel, which I first configured to use bfs at 1000Hz made it use my cpu at 50% at 600MHz. Not really terribly slowing down my system, but <10% are usual. Then I went down to 864Hz, which is recommended by the zen devs and the cpu usage went down to about 35%. I think I'll recompile it with cfs to see if bfs is the evil one here. Other players, like mplayer, behave like before.

cerbie · 2009-09-12 16:38:14

pawels64 wrote:

@Bogart
I have no idea However, this change will be in mainline tree (I don't know if it will be backported). This change should solve latency problems which some people can suffer from. It seems bfs can go to retirement now (very early ).

That's a fix to one small bug. They can want one scheduler to rule them all, but it keeps not working right. The solution, which could retire BFS, would be to allow selecting of scheduler as a config option, either fully hot, at boot-time, or by config, in the mainline kernel. I really don't get why there has constantly been such resistance to the idea.

@Army: what is the supposed benefit of 864Hz? Google only helps me find, "it's the best!" or "it's wrong, and you are wrong, now go away."

Also, after about the 3rd compile (I keep forgetting B/G ralink drivers have been put in experimental and defaulted to off), the .30 BFS seems to act like the realtime kernels did. Yay. Now to go mess with .31, because it's there...

brisbin33 · 2009-09-12 17:43:39

Renan Birck wrote:

[...]had to struggle with pacman to get my old kernel back[...].

you're doing it wrong . they can be installed side by side with a menu.lst entry for each. makes for much more conventient testing.

though i do have -Rd/-S nvidia/nvidia-bfs in my case, but even thats no real bother.

Arch Linux

#101 2009-09-10 21:09:10

Re: omg! Con Kolivas is back!

#102 2009-09-10 21:30:52

Re: omg! Con Kolivas is back!

#103 2009-09-10 22:17:11

Re: omg! Con Kolivas is back!

#104 2009-09-10 22:31:11

Re: omg! Con Kolivas is back!

#105 2009-09-11 07:57:38

Re: omg! Con Kolivas is back!

#106 2009-09-11 14:35:37

Re: omg! Con Kolivas is back!

#107 2009-09-11 15:05:27

Re: omg! Con Kolivas is back!

#108 2009-09-11 17:00:25

Re: omg! Con Kolivas is back!

#109 2009-09-11 17:01:52

Re: omg! Con Kolivas is back!

#110 2009-09-11 19:27:05

Re: omg! Con Kolivas is back!

#111 2009-09-11 19:28:54

Re: omg! Con Kolivas is back!

#112 2009-09-11 19:41:04

Re: omg! Con Kolivas is back!

#113 2009-09-11 19:44:03

Re: omg! Con Kolivas is back!

#114 2009-09-11 21:03:55

Re: omg! Con Kolivas is back!

#115 2009-09-12 01:30:36

Re: omg! Con Kolivas is back!

#116 2009-09-12 01:52:52

Re: omg! Con Kolivas is back!

#117 2009-09-12 02:23:23

Re: omg! Con Kolivas is back!

#118 2009-09-12 03:02:51

Re: omg! Con Kolivas is back!

#119 2009-09-12 04:09:05

Re: omg! Con Kolivas is back!

#120 2009-09-12 09:31:16

Re: omg! Con Kolivas is back!

#121 2009-09-12 09:39:45

Re: omg! Con Kolivas is back!

#122 2009-09-12 13:57:23

Re: omg! Con Kolivas is back!

#123 2009-09-12 14:25:27

Re: omg! Con Kolivas is back!

#124 2009-09-12 16:38:14

Re: omg! Con Kolivas is back!

#125 2009-09-12 17:43:39

Re: omg! Con Kolivas is back!

Board footer