Con Kolivas BFS sched

[Licaon_kter]

:: Code ::

root@mindeba:/home/licaon/ > cat /boot/config-2.6.35-0.dmz.2-liquorix-amd64 | grep BFS 
# CONFIG_SCHED_BFS is not set

Any reason the BFS sched patch is included yet not compiled ?

10x

[damentz]

Yes, it has a hard to find suspend bug that effects h2 (author of smxi, inxi, sgfxi, etc...) and breaks the initialization of the nvidia graphics driver for the gt 330m on my laptop. Fortunately, CFS has come a long way since 2.6.32 where the first attempts at fixing bugs highlighted by BFS were committed. Now it performs as equal or better than BFS on the desktop.

In order to compensate, I reduced the maximum runtime slice per task to 3ms from 6ms (or from 5ms on 2.6.34 and less) to provide faster desktop responsiveness that most users expect. This is a value also suggested by Con to tune BFS for very low latency desktop environments.

Now, to play devil's advocate, there were a lot of changes in 2.6.35 that really bricked BFS for suspend to ram. It took Con unusual amount of time to create an official port BFS in order to fix those bugs. It's likely that in the process the hard to reproduce suspend anomalies are fixed. I'll test BFS again and see if the problems I was first experiencing still exist, and if not, probably use BFS for the process scheduler going forward.

Thanks for the reminder!

[Licaon_kter]

:: damentz wrote ::

It's likely that in the process the hard to reproduce suspend anomalies are fixed. I'll test BFS again and see if the problems I was first experiencing still exist, and if not, probably use BFS for the process scheduler going forward.

Oh, great then. 10x

Cheers!

[Licaon_kter]

Any news?

[damentz]

Ya, it's still not fixed - nvidia's graphics driver is troublesome with my card, the GT 330M. It takes about 10 boots to let the driver initialize my card correctly to rendering things at its fastest speed.

[Licaon_kter]

:: damentz wrote ::

Ya, it's still not fixed - nvidia's graphics driver is troublesome with my card, the GT 330M. It takes about 10 boots to let the driver initialize my card correctly to rendering things at its fastest speed.

oh, i saw your posts on nvnews.net.
did you get this: theinvisiblethings.blogspot.com/2010/08/skeletons-hidden-in-linux-closet.html backported too?

/LE:
the 2.6.35-3.dmz.1-liquorix-amd64 changelog is wrong?

:: Quote ::

linux-liquorix-2.6 (2.6.35-10) unstable; urgency=low

* 2.6.35.3
* updated configuration
- using CFQ instead of BFQ
* update ABI to 3.dmz.1

-- Steven Barrett <damentz@gmail.com> Sun, 22 Aug 2010 01:24:09 -0500

but

:: Quote ::

cat config-2.6.35-3.dmz.1-liquorix-amd64 | grep SCHED
...
CONFIG_DEFAULT_IOSCHED="bfq"
...

[Anonymo]

I didn't see this patch on the website

[damentz]

Oops, didn't catch that when I was updating the configuration. Next update will be on CFQ.

And regarding Anonymo, I just uploaded the new patches - sorry for the delay.

[Licaon_kter]

:: damentz wrote ::

Ya, it's still not fixed - nvidia's graphics driver is troublesome with my card, the GT 330M. It takes about 10 boots to let the driver initialize my card correctly to rendering things at its fastest speed.

version 256.52 ( www.nvnews.net/vbulletin/showthread.php?p=2309077 ) is out ;)

[ulenrich]

- Using this new patch:
ck.kolivas.org/patches/bfs/2.6.35/
- And this my new .config optimized for BFS on Intel-Core2Duo:
sidux.com/index.php?name=PNphpBB2&file=viewtopic&p=164873#164873

I do have a wonderful runing BFS-patched Kernel-2.6.35 :)
Any suggestion to even better configure?