Politics and kernel schedulers

karol · 2011-06-08 00:04:47

redden0t8 wrote:

Annecdotal evidence from end users seems to ubiquitously show BFS provides a better experience. It would be really interesting to see a blind study on this.

It's about Android so it may not mater:
http://lwn.net/Articles/389230/

swetland wrote:

It was an experimental branch. Some folks reported better experiences on some hardware, but blind testing on Droid and NexusOne did not show a user visible performance difference and we opted to stick with the stock scheduler for the our 2.6.32/Froyo efforts.

bwat47 · 2011-06-08 01:26:59

sonay wrote:

Is BFS really better for desktops? If so, I am planning to use it on my netbook, would it worth the trouble?

I'm using kernel26-ck with bfs and bqs on my laptop, its great! I haven't done any scientific benchmarking though.

graysky · 2011-06-08 19:23:10

sonay wrote:

Is BFS really better for desktops? If so, I am planning to use it on my netbook, would it worth the trouble?

It's no trouble. There is an unofficial repo you can use (see my sig). BFS is a terrific scheduler.

sonay · 2011-06-09 17:57:30

Is the binary repo still active?

I get the following error:

error: failed retrieving file 'kernel26-ck.db' from home.comcast.net : No address record
error: failed to update kernel26-ck (No address record)

I am also wondering if there's a rtl8192se module.

karol · 2011-06-09 18:03:05

sonay wrote:

Is the binary repo still active?
I get the following error:
error: failed retrieving file 'kernel26-ck.db' from home.comcast.net : No address record
error: failed to update kernel26-ck (No address record)
I am also wondering if there's a rtl8192se module.

https://bbs.archlinux.org/viewtopic.php?id=114204

Post questions regarding the repo here https://bbs.archlinux.org/viewtopic.php?id=111715

Last edited by karol (2011-06-09 18:06:14)

sonay · 2011-06-09 18:12:47

karol wrote:

https://bbs.archlinux.org/viewtopic.php?id=111715

OK, thanks.

Sorry for dirtying the forum, I just followed from the wiki.I couldn't find a delete message button, maybe I am blind

Last edited by sonay (2011-06-09 18:22:04)

karol · 2011-06-09 18:16:54

[karol@black ~]$ sudo pacman -Syy
:: Synchronizing package databases...
 kernel26-ck                                    4,7K   28,2K/s 00:00:00 [########################################] 100%
...

Seems fine from here.

Last edited by karol (2011-06-09 18:19:08)

graysky · 2011-06-09 18:17:51

@sonay - the server is up just fine. Your entry for pacman.conf looks fine. Post the output of the following, but please do it in the repo help thread which is here.

$ pacman -Syy

jnguyen · 2011-06-09 18:18:46

sonay wrote:

I am also wondering if there's a rtl8192se module.

rtl8192se module is in AUR, but will be included upstream in the next kernel release (3.0).

sonay · 2011-06-09 18:35:46

jnguyen wrote:

...will be included upstream in the next kernel release (3.0).

Niiiice

I meant for the kernel26-ck, as there are some modules for other devices and non for mine in the wiki list. Will the Aur package for vanilla kernel work?

Misfit138 · 2011-06-11 00:49:28

As has been stated many times, this is a highly polarized issue which is caught between two personalities; Linus and Con.
From what I have gathered, Con seems like a very reasonable and competent guy who offered a superior scheduler which was rejected by Linus. Soon thereafter, the CFS was implemented into mainline, basically taking ideas from Con and giving him no credit.
This just basically reinforces my perception of Linus as being a bit of an egomaniac and control freak.

Octoploid · 2011-06-11 19:50:24

Con is a winer, who doesn't know how to handle any kind of criticism.
That was one of the reasons Linus didn't take BFS originally.
And he was right; BFS is still unusable buggy crap today, while the kernel
scheduler is working very fine and is rock stable.

Last edited by Octoploid (2011-06-11 19:53:29)

Gusar · 2011-06-11 20:01:15

Octoploid wrote:

BFS is still unusable buggy crap today, while the kernel

That's a very loaded statement. And it's tone does not put you in a position that would allow you to call people names.

bwat47 · 2011-06-11 20:13:40

Octoploid wrote:

Con is a winer, who doesn't know how to handle any kind of criticism.
That was one of the reasons Linus didn't take BFS originally.
And he was right; BFS is still unusable buggy crap today, while the kernel
scheduler is working very fine and is rock stable.

I haven't had a single problem with bfs. Where the hell do you get "unusable" from? Don't make such baseless allegations.

Last edited by bwat47 (2011-06-11 20:14:19)

graysky · 2011-06-11 20:19:01

octoploid = troll
ignore him

jasonwryan · 2011-06-11 20:21:18

This thread is supposed to be about schedulers not about personalities -- if you insist on dragging it down to that level, the thread will be closed.

Similarly, please show respect for each other and address the issue at hand: https://wiki.archlinux.org/index.php/Fo … ther_Users

fsckd · 2011-06-12 01:52:24

Mainline scheduler has been improving. One year ago it was quite crap. Now it's not too shabby. I'd like to think of BFS as a motivator for this.

Gusar · 2011-06-12 10:02:17

fsckd wrote:

I'd like to think of BFS as a motivator for this.

Definitely. It was a test with the x264 encoder that showed mainline blowing huge donkey balls compared to BFS back then. Using that testcase, mainline was then fixed. And other small improvements followed. Nowadays the two schedulers are much closer. But back then I saw a huge improvement with BFS especially on a crappy Atom N270 processor, so now I'm sticking with it even on other processors.

brebs · 2011-06-12 10:18:19

BFS is needed for:

man schedtool wrote:

SCHED_ISO [ patch needed ] SCHED_ISO was designed to give users a SCHED_RR-similar class. To quote Con Kolivas: "This is a non-expiring scheduler policy designed to guarantee a timeslice within a reasonable latency while preventing starvation. Good for gaming, video at the limits of hardware, video capture etc."
SCHED_IDLEPRIO [ patch needed ] SCHED_IDLEPRIO is similar to SCHED_BATCH, but was explicitly designed to consume only the time the CPU is idle. No interactive boosting is done. If you used SCHED_BATCH in the -ck kernels this is what you want since 2.6.16

That IDLEPRIO is good for running e.g. transmission in the background.

Arch Linux

#26 2011-06-08 00:04:47

Re: Politics and kernel schedulers

#27 2011-06-08 01:26:59

Re: Politics and kernel schedulers

#28 2011-06-08 19:23:10

Re: Politics and kernel schedulers

#29 2011-06-09 17:57:30

Re: Politics and kernel schedulers

#30 2011-06-09 18:03:05

Re: Politics and kernel schedulers

#31 2011-06-09 18:12:47

Re: Politics and kernel schedulers

#32 2011-06-09 18:16:54

Re: Politics and kernel schedulers

#33 2011-06-09 18:17:51

Re: Politics and kernel schedulers

#34 2011-06-09 18:18:46

Re: Politics and kernel schedulers

#35 2011-06-09 18:35:46

Re: Politics and kernel schedulers

#36 2011-06-11 00:49:28

Re: Politics and kernel schedulers

#37 2011-06-11 19:50:24

Re: Politics and kernel schedulers

#38 2011-06-11 20:01:15

Re: Politics and kernel schedulers

#39 2011-06-11 20:13:40

Re: Politics and kernel schedulers

#40 2011-06-11 20:19:01

Re: Politics and kernel schedulers

#41 2011-06-11 20:21:18

Re: Politics and kernel schedulers

#42 2011-06-12 01:52:24

Re: Politics and kernel schedulers

#43 2011-06-12 10:02:17

Re: Politics and kernel schedulers

#44 2011-06-12 10:18:19

Re: Politics and kernel schedulers

Board footer