View previous topic :: View next topic |
Author |
Message |
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Tue Oct 18, 2005 10:24 am Post subject: [ACID-SOURCES] 2.6.14-rc4-acid1[__DEPRECATED__] |
|
|
NEW RELEASE: 2.6.14-acid1
[__DEPRECATED__]acid-sources-2.6.14-rc4-r1 "The Beginning"
I would like to introduce you to something i have been thinking about
for a long time: a new patchset based upon a sane featureset and
taking advantage of the proven and still well maintained
O(1) scheduler from Ingo Molnar.
This release is a proof of concept, the solid performance and stability i have
been experiencing during the testing as well as absence of breakage-suspicious
cutting edge noone-knows-what-it-really-does patches makes it at least worth
YOUR attention.
Disclaimer
This patchset is in no way affiliated with the Gentoo Foundation and so it's
use does not imply any more or less official support from the gentoo developers.
Do NOT, EVER, post bugs regarding a kernel patched to acid-sources to
_bugs.gentoo.org_. Post your problems in this thread instead.
I would also appreciate that you do not open new topics about problems
related to this kernel, i will be able to help you quicker and more efficiently
in this thread.
What's in it?
Code: |
#
# Performance-tuning related
#
broken_out/sched-tuning.diff //Con Kolivas and others
broken_out/Con_s-tweaks.diff //Con Kolivas
broken_out/as_iosched_tuning.diff //Nick Piggin
broken_out/readahead-tune.diff //Suse
broken_out/swaptoken-tuning.diff //Rik van Riel
broken_out/swap-prefetch-from-rc4-mm1_config-default-n.diff //Con Kolivas
#
# Must have
#
broken_out/vesafb-tng-1.0-rc1-r1-and-fbsplash-0.9.2-r4-for-2.6.14-rc4.diff //Spock
broken_out/software-suspend-2.2-rc8-for-2.6.14-rc4.diff //Nigel Cunningham
broken_out/reiser4-from-2.6.14-rc4-mm1.diff //Hans Reiser\Namesys
broken_out/svgalib_helper-1.9.21.diff //Brion Vibber and others
#
# Some people need these too
#
broken_out/shfs-0.35_for_2.6.14-rc4.diff //Zemljanka core team
broken_out/squashfs_2.2-r2.diff //Phillip Lougher
broken_out/unionfs-20051014-1448.diff //www.fsl.cs.sunysb.edu
broken_out/usbhid-readd-kconfig.diff //Mikkel Krautz
broken_out/daconfig-2.3.0.diff // DaMouse
#
# Minor additions (especially libata and netdev git-trees)
# Read patchnames
#
broken_out/git_updates_and_few_fixes.diff //Various
broken_out/iosched_module_refcount_leak_fix-and-def_iosched_submenu.diff // Various->lkml
broken_out/vmalloc_node.diff //Christoph Lameter
broken_out/write_inode_now-write-inode-if-not-bdi_cap_no_writeback.diff //Andrew Morton
broken_out/shutup-unneeded-warnings.diff //Tiger683 (ugly, ugly)
#
# Acid-Sources only
#
broken_out/acid-sources-framebuffer-logo.diff //Tiger683 (ugly, ugly)
broken_out/acidize.diff //Tiger683 (ugly, ugly)
|
And now, even more important part:
LINKS
MD5SUMS
EBUILD
PATCH
BROKEN OUT TARBALL
IMPORTANT UPDATES AND EDITS:
*** I would appreciate if anyone could do a better FB logo, and provide it in png format, thx in advance
** To be continued...
Howto install
Code: |
GENTOO:
1)create directory <OVERLAY_DIR>/sys-kernel/acid-sources (substitute <OVERLAY_DIR> according to overlay path)
2) PROMPT>cd <OVERLAY_DIR>/sys-kernel/acid-sources
3) PROMPT>wget http://www.stud.uni-karlsruhe.de/~uyavl/public/acid-sources/2.6.14-rc4-acid1/acid-sources-2.6.14_rc4-r1.ebuild
4) PROMPT>ebuild acid-sources-2.6.14_rc4-r1.ebuild digest
5) PROMPT>USE="symlink" emerge acid-sources
6) PROMPT>cd /usr/src/linux
7) configure, compile and install your kernel
if in doubt about how to do it, consult the documentation in the kernel source directory
ALL OTHER DISTROS:
1) PROMPT>cd /usr/src
2) PROMPT>wget http://www.kernel.org/pub/linux/kernel/v2.6/testing/linux-2.6.14-rc4.tar.bz2
3) PROMPT>wget http://www.stud.uni-karlsruhe.de/~uyavl/public/acid-sources/2.6.14-rc4-acid1/patch-2.6.14-rc4-acid1.bz2
4) PROMPT>tar -xvjf linux-2.6.14-rc4.tar.bz2 && rm linux-2.6.14-rc4.tar.bz2
5) PROMPT>mv linux-2.6.14-rc4 linux-2.6.14-rc4-acid1 && rm linux && ln -s linux-2.6.14-rc4-acid1 linux
6) PROMPT>bunzip2 patch-2.6.14-rc4-acid1.bz2 && cd linux
7) PROMPT>patch -p1 -i ../patch-2.6.14-rc4-acid1
8) configure, compile and install your kernel
if in doubt about how to do it, consult the documentation in the kernel source directory
|
Have fun...
T _________________ Retired gentoo user
Last edited by Tiger683 on Fri Oct 28, 2005 3:29 pm; edited 2 times in total |
|
Back to top |
|
|
bob_111 Apprentice
Joined: 12 Oct 2004 Posts: 155
|
Posted: Tue Oct 18, 2005 12:00 pm Post subject: |
|
|
Ahh another patchset! Time to put these to the test against nitro-sources!
- bob_111
Edit: Just got em installed. All went good except my dodgy fix that i have to apply to EVERY kernel i try!
Code: | michael@PikleBox ~ $ uname -a
Linux PikleBox 2.6.14-rc4-acid1 #2 SMP Wed Oct 19 08:50:35 EST 2005 x86_64 AMD Athlon(tm) 64 Processor 3200+ AuthenticAMD GNU/Linux |
Only con is that i lost 6fps of my glxgears (nitro 7087, acid 7081) but no biggie!
Last edited by bob_111 on Tue Oct 18, 2005 1:28 pm; edited 1 time in total |
|
Back to top |
|
|
vipernicus Veteran
Joined: 17 Jan 2005 Posts: 1462 Location: Your College IT Dept.
|
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Tue Oct 18, 2005 1:31 pm Post subject: |
|
|
vipernicus wrote: | Code: | broken_out/Con_s-tweaks.diff //Con Kolivas |
What all does this include? |
Con_s-tweaks.diff wrote: |
hz-default_values.patch
hz-no250.patch
pdflush-tweaks.patch
1g_lowmem1_i386.diff
|
Oh, and Code: | broken_out/sched-tuning.diff //Con Kolivas and others |
is not staircase
If more details needed, these are the 13 sched patches in current mm-tree (2.6.14-rc4-mm1)
Mostly targeting SMP machines, but not only... _________________ Retired gentoo user |
|
Back to top |
|
|
bollucks l33t
Joined: 27 Oct 2004 Posts: 606
|
Posted: Tue Oct 18, 2005 11:36 pm Post subject: Re: [ACID-SOURCES] 2.6.14-rc4-acid1 |
|
|
Tiger683 wrote: | I would like to introduce you to something i have been thinking about
for a long time: a new patchset based upon a sane featureset and
taking advantage of the proven and still well maintained
O(1) scheduler from Ingo Molnar.
This release is a proof of concept, the solid performance and stability i have
been experiencing during the testing as well as absence of breakage-suspicious
cutting edge noone-knows-what-it-really-does patches makes it at least worth
YOUR attention. |
Interesting... all -nitro breakage we've seen hasn't actually been the cpu scheduler but other stuff you've included. |
|
Back to top |
|
|
predatorfreak l33t
Joined: 13 Jan 2005 Posts: 708 Location: USA, Michigan.
|
Posted: Wed Oct 19, 2005 12:13 am Post subject: |
|
|
Not to be too negative, but I still see alot of pointless patches in here or things which you don't really need/won't enhance proformance. Also, 1GB lowmem isn't really going to improve proformance that much, according to Con it is supposed to work around some overhead with using high mem for 1GB, which causes bad network proformance (If I remember correctly). _________________ System: predatorbox
Distro: Arch Linux x86_64
Current projects: blackhole, convmedia and anything else I cook up. |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Wed Oct 19, 2005 12:23 am Post subject: |
|
|
predatorfreak wrote: | according to Con it is supposed to work around some overhead with using high mem for 1GB (..) |
Exactly.. See my point?
Even if you don't have exactly 1G of RAM does't mean that I don't either. Well, actually i do, just like many people in here...
bollucks wrote: | -nitro breakage (...) stuff you've included (...) |
Hmm, tell it the laptop users who demand software suspend 2, the win4lin freaks and developers of proprietary modules (eg. ati) who dont keep up with current kernels. Other than that, the ammount of -mm patches was being decreased step-by-step as we proceeded toward getting stable, and though many did stay anyway. See example: 2.6.12-nitro5 _________________ Retired gentoo user
Last edited by Tiger683 on Wed Oct 19, 2005 1:43 am; edited 1 time in total |
|
Back to top |
|
|
predatorfreak l33t
Joined: 13 Jan 2005 Posts: 708 Location: USA, Michigan.
|
Posted: Wed Oct 19, 2005 12:55 am Post subject: |
|
|
Tiger683 wrote: | predatorfreak wrote: | according to Con it is supposed to work around some overhead with using high mem for 1GB (..) |
Exactly.. See my point?
Even if you don't have exactly 1G of RAM does't mean that I don't either. Well, actually i do, just like many people in here...
bollucks wrote: | -nitro breakage (...) stuff you've included (...) |
Hmm, tell it the laptop users who demand software suspend 2, the win4lin freaks and developers of proprietary modules (eg. ati) who dont keep up with current kernels. Other than that, the ammount of -mm patches was being decreased step-by-step as we proceeded toward getting stable, and though many did stay anyway. See example: 2.6.12-nitro5 |
The overhead is ignorable to some and it might not even appear for others. _________________ System: predatorbox
Distro: Arch Linux x86_64
Current projects: blackhole, convmedia and anything else I cook up. |
|
Back to top |
|
|
bollucks l33t
Joined: 27 Oct 2004 Posts: 606
|
Posted: Wed Oct 19, 2005 1:03 am Post subject: |
|
|
Tiger683 wrote: | bollucks wrote: | -nitro breakage (...) stuff you've included (...) |
Hmm, tell it the laptop users who demand software suspend 2, the win4lin freaks and developers of proprietary modules (eg. ati) who dont keep up with current kernels. Other than that, the ammount of -mm patches was being decreased step-by-step as we proceeded toward getting stable, and though many did stay anyway. See example: 2.6.12-nitro5 |
Actually my point was you didn't include staircase and its extra policies which have been stable. |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Wed Oct 19, 2005 1:48 am Post subject: |
|
|
predatorfreak wrote: |
The overhead is ignorable to some and it might not even appear for others. |
Glad to hear you think Con's work is redundant and, well, as you indirectly said useless.
If you want to further discuss it, open a topic in OffTheWall and send me an invitation, instead of polluting this one with negative comments...
@bollucks:
Sorry, i just got a little over the line... _________________ Retired gentoo user |
|
Back to top |
|
|
predatorfreak l33t
Joined: 13 Jan 2005 Posts: 708 Location: USA, Michigan.
|
Posted: Wed Oct 19, 2005 2:27 am Post subject: |
|
|
Tiger683 wrote: | predatorfreak wrote: |
The overhead is ignorable to some and it might not even appear for others. |
Glad to hear you think Con's work is redundant and, well, as you indirectly said useless.
If you want to further discuss it, open a topic in OffTheWall and send me an invitation, instead of polluting this one with negative comments...
@bollucks:
Sorry, i just got a little over the line... |
Sigh, it's not useless or redundant, but most users won't expierance the overhead. None the less, the overhead is there and when you do get that overhead, having this patch available would be a great help. Also, my comments are not trying to be negative, but to be helpful so you can produce a better and cleaner patchset.
Edit +not _________________ System: predatorbox
Distro: Arch Linux x86_64
Current projects: blackhole, convmedia and anything else I cook up. |
|
Back to top |
|
|
cheater1034 Veteran
Joined: 09 Sep 2004 Posts: 1558
|
Posted: Wed Oct 19, 2005 2:33 am Post subject: |
|
|
I love nitro-sources myself
----
Looks interesting, how does it match up against nitro?
And anyway, good work, this has some nice necessary patches that I could use. _________________ IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land. |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Wed Oct 19, 2005 12:21 pm Post subject: |
|
|
For me it performs slightly better ATM, give it a shot if you like
I need some Inspiration about any new features, as the rc kernel is a good oportunity to test stuff for next stable release.
cheers
T _________________ Retired gentoo user |
|
Back to top |
|
|
_sys/sid Guru
Joined: 27 Aug 2004 Posts: 346 Location: Asola (Mantova)
|
Posted: Wed Oct 19, 2005 4:56 pm Post subject: |
|
|
Code: |
laurien@nebula [~] $ uname -a
Linux nebula 2.6.14-rc4-acid1 #1 SMP Wed Oct 19 17:03:10 CEST 2005 i686 Intel(R) Pentium(R) 4 CPU 2.60GHz GenuineIntel GNU/Linux
|
Code: |
laurien@nebula [~] $ cat /proc/version
Linux version 2.6.14-rc4-acid1 (root@nebula) (gcc version 3.3.6 (Gentoo 3.3.6, ssp-3.3.6-1.0, pie-8.7.8)) #1 SMP Wed Oct 19 17:03:10 CEST 2005
|
_________________ http://th30z.netsons.org/
Last edited by _sys/sid on Wed Oct 19, 2005 5:00 pm; edited 1 time in total |
|
Back to top |
|
|
longodj Tux's lil' helper
Joined: 29 Aug 2003 Posts: 79 Location: New Jersey
|
Posted: Wed Oct 19, 2005 4:58 pm Post subject: e1000 error |
|
|
Hey, I really like your idea
I tried to install your kernel and got an error with the Intel PRO/1000 driver ( e1000 ). It failed to compile that driver at all... Anyway, my genkernel log can be found at http://cloudnine.rh.rit.edu/~sandman/genkernel.log
After disabling e1000 in the config ( I have nforce networking ), the kernel compiled fine. I'm just about to reboot to try it out I'm trying your kernel because on nitro I'm experiencing some instability relating I think to reiser4. I don't know if it is caused by nitro but my computer randomly dies. I don't mean freezes, I mean dies. I am assuming it's fs related because when this happens...say I'm running a terminal and I use ls, it will hang. From there, I can use a web browser for a few seconds, but I can't open anything. Then after a little while everything will just stop, and I have to reboot. I hope this fixes it
Thanks alot,
Dave _________________ Linux dreammachine 2.6.16-rc3-nitro1 #1 PREEMPT Thu Feb 23 15:56:10 EST 2006 x86_64 AMD Athlon(tm) 64 Processor 3800+ GNU/Linux
1Gb RAM
Quadro FX 1300
2 x 36Gb Raptors
1 x 400Gb
Plextor Slot-Load 16X DVD+/-RW
(Linux Friendly FSF Donor) |
|
Back to top |
|
|
longodj Tux's lil' helper
Joined: 29 Aug 2003 Posts: 79 Location: New Jersey
|
Posted: Wed Oct 19, 2005 5:02 pm Post subject: |
|
|
Sweet, worked first shot, only had to remerge nvidia-kernel as usual. Good job, feel faster too. I REALLY like the logo.
Linux localhost 2.6.14-rc4-acid1 #1 PREEMPT Wed Oct 19 08:39:13 EDT 2005 x86_64 AMD Athlon(tm) 64 Processor 3800+ AuthenticAMD GNU/Linux _________________ Linux dreammachine 2.6.16-rc3-nitro1 #1 PREEMPT Thu Feb 23 15:56:10 EST 2006 x86_64 AMD Athlon(tm) 64 Processor 3800+ GNU/Linux
1Gb RAM
Quadro FX 1300
2 x 36Gb Raptors
1 x 400Gb
Plextor Slot-Load 16X DVD+/-RW
(Linux Friendly FSF Donor) |
|
Back to top |
|
|
SchAmane Tux's lil' helper
Joined: 11 Sep 2004 Posts: 107
|
Posted: Wed Oct 19, 2005 5:16 pm Post subject: |
|
|
Nvidia,Reiser4 works well here. Feels better than 2.6.14-rc3-nitro3 and a lot smoother.
Copying from CD-ROM takes half of my CPU even if i compile world but whole system is very responsive untill it. Thats much better than on nitro.
Naturaly if i set DMA on, it works even better |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Wed Oct 19, 2005 5:21 pm Post subject: |
|
|
// world's shortest meaningful post _________________ Retired gentoo user |
|
Back to top |
|
|
longodj Tux's lil' helper
Joined: 29 Aug 2003 Posts: 79 Location: New Jersey
|
Posted: Wed Oct 19, 2005 5:34 pm Post subject: |
|
|
Yea...dma rocks
Code: | hdparm -d1 -A1 -M1 /dev/sdb |
I'm not in gentoo right now but it basically increased from 2400Mb/s to 2800Mb/s _________________ Linux dreammachine 2.6.16-rc3-nitro1 #1 PREEMPT Thu Feb 23 15:56:10 EST 2006 x86_64 AMD Athlon(tm) 64 Processor 3800+ GNU/Linux
1Gb RAM
Quadro FX 1300
2 x 36Gb Raptors
1 x 400Gb
Plextor Slot-Load 16X DVD+/-RW
(Linux Friendly FSF Donor) |
|
Back to top |
|
|
cheater1034 Veteran
Joined: 09 Sep 2004 Posts: 1558
|
Posted: Wed Oct 19, 2005 6:56 pm Post subject: |
|
|
I'll definitely have to try this out some time.
If you read my sig, you'd know I'm building a gentoo minimal livecd with reiser4progs/nitro-sources and a few other progs.
Actually I'm almost to making the squashfs image as we speak (I'm in links)
I'm going to test nitro and acid, depending on which one I like better, I'll use it in my next release of the livecd --- My 2005.1 x86 Conrad gentoo install cd should be done soon, if acid performs better than nitro I'll include the patchset over nitro on the livecd in the next release (probally 2005.2, or whatever next gentoo release is)
I will try it as soon as I build this livecd and install gentoo with reiser4 _________________ IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land. |
|
Back to top |
|
|
babykart Guru
Joined: 08 Oct 2004 Posts: 415
|
Posted: Wed Oct 19, 2005 7:54 pm Post subject: |
|
|
I don't understand why but this kernel doesn't detect my second raptor...
well, for the moment, I stay on the last nitro... _________________ >> Gentoo-FR <<
----- |
|
Back to top |
|
|
thubble Tux's lil' helper
Joined: 16 May 2003 Posts: 119 Location: Winnipeg, Canada
|
Posted: Wed Oct 19, 2005 7:57 pm Post subject: |
|
|
predatorfreak wrote: | Tiger683 wrote: | predatorfreak wrote: |
The overhead is ignorable to some and it might not even appear for others. |
Glad to hear you think Con's work is redundant and, well, as you indirectly said useless.
If you want to further discuss it, open a topic in OffTheWall and send me an invitation, instead of polluting this one with negative comments...
@bollucks:
Sorry, i just got a little over the line... |
Sigh, it's not useless or redundant, but most users won't expierance the overhead. None the less, the overhead is there and when you do get that overhead, having this patch available would be a great help. Also, my comments are not trying to be negative, but to be helpful so you can produce a better and cleaner patchset.
Edit +not |
I agree. The thing about the 1GB patch is that it affects only one file, and if you disable it in .config there is literally no way it can affect anything. I like having the full 1024MB instead of <900, although I don't do anything that memory-intensive so I rarely notice the difference.
I'm going to try this kernel now, it's been a while since I've tried anything O(1)-based. |
|
Back to top |
|
|
SchAmane Tux's lil' helper
Joined: 11 Sep 2004 Posts: 107
|
Posted: Wed Oct 19, 2005 8:24 pm Post subject: |
|
|
Another interesting thing about this kernel:
if you start something like compiling programms. You get 100% CPU usage (as usual). But if you start top (or htop) and sort you running processes on CPU usage, it wouldnt be a compiler that takes most CPU time And it seems like it take little bit more time to compile things than on nitro, but i have not pruf that realy now. Maybe somebody have same behavior?
Addition:
Here is a picture about what i am speaking:
http://screens.myeburg.net/top_for_acid.png
Interesting that even on this very hight system load all applications are good responsible |
|
Back to top |
|
|
satanskin Guru
Joined: 25 Apr 2005 Posts: 353
|
Posted: Wed Oct 19, 2005 10:42 pm Post subject: |
|
|
Damnit! Now I have to try this out! You all make it sound so good! Just curious, but what are you all using as your default I/O scheduler? |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Wed Oct 19, 2005 11:12 pm Post subject: |
|
|
anticipatory 4 ever !
seriously, thats why i put those anticipatory tune patches from mm in it :p _________________ Retired gentoo user |
|
Back to top |
|
|
|