View previous topic :: View next topic |
Author |
Message |
vipernicus Veteran
Joined: 17 Jan 2005 Posts: 1462 Location: Your College IT Dept.
|
|
Back to top |
|
|
DocterD Tux's lil' helper
Joined: 15 May 2004 Posts: 129
|
Posted: Thu Apr 26, 2007 7:32 pm Post subject: |
|
|
could you merge dynticks for x86_64 in your next release? |
|
Back to top |
|
|
vipernicus Veteran
Joined: 17 Jan 2005 Posts: 1462 Location: Your College IT Dept.
|
|
Back to top |
|
|
vipernicus Veteran
Joined: 17 Jan 2005 Posts: 1462 Location: Your College IT Dept.
|
Posted: Thu Apr 26, 2007 7:46 pm Post subject: |
|
|
immudium wrote: | vipernicus wrote: | If I can get 2 or 3 more votes, I'll revert back to SD for 2.6.21. |
Give us a little more time with CFS before you decide. I'm pleased with it so far, but it's way too early to pass judgment. I could care less about higher temps, if any. My computer is here to do work, not sit there cooling itself off.
Edit:
Now that I've had to play around with it some more, I'm even more impressed with CFS. I typically run a mix of Java/Swing apps and Wine apps such as newsbin with a heavy dose of multimedia, mp3, x264. With Viper9/SD, wine/newsbin would become unresponsive after a couple of minutes like it was being starved. Also the audio of my mp3 and/or video would "hiccup" when context switching between windows in my java apps. With CFS and the same nice priority of -10, everything runs very smoothly. No glitches in audio, no unresponsive wine apps. |
Try Viper9/SD with a prio boost of -19. Let me know if it reacts the same as CFS at -10. _________________ Viper-Sources Maintainer || nesl247 Projects || vipernicus.org blog |
|
Back to top |
|
|
RobertDavid Apprentice
Joined: 08 Aug 2005 Posts: 271 Location: Czech Republic
|
Posted: Thu Apr 26, 2007 7:54 pm Post subject: |
|
|
immudium wrote: | vipernicus wrote: | If I can get 2 or 3 more votes, I'll revert back to SD for 2.6.21. |
Give us a little more time with CFS before you decide. I'm pleased with it so far, but it's way too early to pass judgment. I could care less about higher temps, if any. My computer is here to do work, not sit there cooling itself off.
Edit:
Now that I've had to play around with it some more, I'm even more impressed with CFS. I typically run a mix of Java/Swing apps and Wine apps such as newsbin with a heavy dose of multimedia, mp3, x264. With Viper9/SD, wine/newsbin would become unresponsive after a couple of minutes like it was being starved. Also the audio of my mp3 and/or video would "hiccup" when context switching between windows in my java apps. With CFS and the same nice priority of -10, everything runs very smoothly. No glitches in audio, no unresponsive wine apps. |
Agrea with you, Im also here not to keep my computer cool, but: first the computer is significantly louder (mainly in the night) and second if you run 24/7 you also keep the computer to consume as little electricity as you can..
Second thing: I have never seen audio jumping/glitching problems on linux, maybe you have something bad configured.. (On my 600MHz notebook I sometimes open many programs and burn the cpu to 100% but no audio problems), maybe some wine bug (I can only see that few programs I run wine just "sleep" for a while when they get 100% CPU but it not affect native linux programs simulatenously running) _________________ Arch & Fluxbox & 2.6.24-rc-zen!!!!
robertek.brevnov.net |
|
Back to top |
|
|
vipernicus Veteran
Joined: 17 Jan 2005 Posts: 1462 Location: Your College IT Dept.
|
Posted: Thu Apr 26, 2007 7:59 pm Post subject: |
|
|
RobertDavid wrote: | Looks nice, work fine, but I think CFS is not stuff for us with ATHCOOL activated, because it works against it, and its not good for long time usage (for me I have common 7+ days runtime), with SD I have 40-42'C, with CFS 50-51'C and it is like athcool disabled. And may or may not I dont a see a interactivity increase, (maybe Im on hw peek with SD, or CFS),, but I will continue to test that. |
Do you think this might be because I left tickless out of this kernel. I'd think that would make a bit of a difference. _________________ Viper-Sources Maintainer || nesl247 Projects || vipernicus.org blog |
|
Back to top |
|
|
RobertDavid Apprentice
Joined: 08 Aug 2005 Posts: 271 Location: Czech Republic
|
Posted: Thu Apr 26, 2007 8:47 pm Post subject: |
|
|
May be, but I didnt have any bad results w/o tickless..?!? _________________ Arch & Fluxbox & 2.6.24-rc-zen!!!!
robertek.brevnov.net |
|
Back to top |
|
|
vipernicus Veteran
Joined: 17 Jan 2005 Posts: 1462 Location: Your College IT Dept.
|
Posted: Thu Apr 26, 2007 8:54 pm Post subject: |
|
|
RobertDavid wrote: | May be, but I didnt have any bad results w/o tickless..?!? |
You didn't have any bad results without tickless? I don't understand. Do you mean you didn't run -viper9 with tickless? _________________ Viper-Sources Maintainer || nesl247 Projects || vipernicus.org blog |
|
Back to top |
|
|
RobertDavid Apprentice
Joined: 08 Aug 2005 Posts: 271 Location: Czech Republic
|
Posted: Thu Apr 26, 2007 9:01 pm Post subject: |
|
|
viper9 with tickless, but I thing that I tryed also viper8 w/o,,, So I have adopted the patch from viper9 and now compiling...
EDIT: it work and is better, now the temperature goes down to ~44'C, so some 7-8'C less then before,, thanks for advice:)
And the responsivity is realy improved with CFS... (also I feel now more responsive than before w/o tickless, but it is subjective) _________________ Arch & Fluxbox & 2.6.24-rc-zen!!!!
robertek.brevnov.net |
|
Back to top |
|
|
vipernicus Veteran
Joined: 17 Jan 2005 Posts: 1462 Location: Your College IT Dept.
|
Posted: Fri Apr 27, 2007 12:20 am Post subject: |
|
|
RobertDavid wrote: | And the responsivity is realy improved with CFS... (also I feel now more responsive than before w/o tickless, but it is subjective) |
Has your vote changed?
Edit: Well, according to the votes on vipernicus.org, SD wins. So expect 2.6.21-viper1 to be SD with boost, and also 2.6.21-nicus1 to be CFSv6. _________________ Viper-Sources Maintainer || nesl247 Projects || vipernicus.org blog |
|
Back to top |
|
|
immudium Guru
Joined: 12 Oct 2004 Posts: 332 Location: Utah
|
Posted: Fri Apr 27, 2007 1:42 am Post subject: |
|
|
vipernicus wrote: | immudium wrote: | vipernicus wrote: | If I can get 2 or 3 more votes, I'll revert back to SD for 2.6.21. |
Give us a little more time with CFS before you decide. I'm pleased with it so far, but it's way too early to pass judgment. I could care less about higher temps, if any. My computer is here to do work, not sit there cooling itself off.
Edit:
Now that I've had to play around with it some more, I'm even more impressed with CFS. I typically run a mix of Java/Swing apps and Wine apps such as newsbin with a heavy dose of multimedia, mp3, x264. With Viper9/SD, wine/newsbin would become unresponsive after a couple of minutes like it was being starved. Also the audio of my mp3 and/or video would "hiccup" when context switching between windows in my java apps. With CFS and the same nice priority of -10, everything runs very smoothly. No glitches in audio, no unresponsive wine apps. |
Try Viper9/SD with a prio boost of -19. Let me know if it reacts the same as CFS at -10. |
Well, I've gone back to viper9 and ran my typical desktop environment on several different SD Priority Boost values (-19, -10, -7, 0). My typical desktop environment consists of Eclipse, VLC Media player, several java apps running from eclipse and newsbin running in wine.
With all SD Boost values less than 0, audio glitches occur when switching between java windows and apps and newsbin in wine becomes unresponsive after about 2-3 minutes.
With SD Boost at 0, audio glitches still occur but less frequently. Wine/newsbin works much better and never becomes unresponsive or stops working.
With CFS at -10, I never got any audio glitches and never had any problems with wine apps after about 8 hours of use today.
I'll be sad to see CFS go next release. We hardly knew ye! |
|
Back to top |
|
|
bollucks l33t
Joined: 27 Oct 2004 Posts: 606
|
|
Back to top |
|
|
immudium Guru
Joined: 12 Oct 2004 Posts: 332 Location: Utah
|
Posted: Fri Apr 27, 2007 6:29 am Post subject: |
|
|
Interesting read. Sounds like he basically came to the opposite conclusion I did about CFS vs SD, but then again he doesn't mention adjusting X priorities or anything, but it's nice to get another point of view with a different set of use cases. And based upon the comments he makes about the startup times of his apps being SLOW, he is in serious need of some hash-style voodoo and a perusal through cheater1034's install guide! |
|
Back to top |
|
|
bollucks l33t
Joined: 27 Oct 2004 Posts: 606
|
Posted: Fri Apr 27, 2007 6:39 am Post subject: |
|
|
Indeed, and the jury on whether we should renice X at all or not is still out, so his test case is very interesting for when the default is to not renice. |
|
Back to top |
|
|
Rion Guru
Joined: 12 Oct 2006 Posts: 383 Location: Minsk, Belarus
|
Posted: Fri Apr 27, 2007 7:27 am Post subject: |
|
|
i've made some more tests with this bug https://forums.gentoo.org/viewtopic-p-4024991.html#4024991
i compile kenrnel with genkernel. and use initramfs-genkernel file for boot splash on kernel initialization.
if i remove initrd from my grub.conf system boots fine.
about suggested patch:
i dont like patches which is only workaround but not solve real problem, so i not tested it anymore. it's better to boot without initrd
PS viper10 _________________ rion-overlay |
|
Back to top |
|
|
darklegion Guru
Joined: 14 Nov 2004 Posts: 468
|
Posted: Fri Apr 27, 2007 10:24 am Post subject: |
|
|
This release seems to be using the bugged reiser4 patch as /bin/true was corrupted after a few hours of use.Use the reiser4 patch from viper4 instead.
This problem probably occurs with 2.6.21-nicus1 as well as it using the same patch. |
|
Back to top |
|
|
vipernicus Veteran
Joined: 17 Jan 2005 Posts: 1462 Location: Your College IT Dept.
|
Posted: Fri Apr 27, 2007 10:36 am Post subject: |
|
|
darklegion wrote: | This release seems to be using the bugged reiser4 patch as /bin/true was corrupted after a few hours of use.Use the reiser4 patch from viper4 instead.
This problem probably occurs with 2.6.21-nicus1 as well as it using the same patch. |
There is no nonbugged reiser4 patch, the fix for the 0'ed files hasn't been released by reiser4 developers yet. _________________ Viper-Sources Maintainer || nesl247 Projects || vipernicus.org blog |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
Rion Guru
Joined: 12 Oct 2006 Posts: 383 Location: Minsk, Belarus
|
Posted: Fri Apr 27, 2007 1:22 pm Post subject: |
|
|
vipernicus wrote: | There is no nonbugged reiser4 patch, the fix for the 0'ed files hasn't been released by reiser4 developers yet. | now understand why it happen on my system. so i'm going to try jfs
but where can i spy on changes in reiser4 code ? _________________ rion-overlay |
|
Back to top |
|
|
darklegion Guru
Joined: 14 Nov 2004 Posts: 468
|
Posted: Sat Apr 28, 2007 5:19 am Post subject: |
|
|
vipernicus wrote: | darklegion wrote: | This release seems to be using the bugged reiser4 patch as /bin/true was corrupted after a few hours of use.Use the reiser4 patch from viper4 instead.
This problem probably occurs with 2.6.21-nicus1 as well as it using the same patch. |
There is no nonbugged reiser4 patch, the fix for the 0'ed files hasn't been released by reiser4 developers yet. |
Are you sure? Its never happened with me on viper4 or dark6 (which I patched in the old viper4 reiser4 patch) and thats after weeks (months?) of use.It has ,however, happened with viper7 and it happened almost immediatly with viper10.Maybe its just a coincidence, though.I will note that booting back into a kernel with the older patch does not reverse the problem, but it doesn't seem to re-occur after emerging coreutils.
I'm currently running 2.6.21-nicus1 with the viper4 patch, with no problems yet.I'll stand corrected if it happens again, though |
|
Back to top |
|
|
kEiNsTeiN Guru
Joined: 29 Apr 2005 Posts: 361 Location: Germany
|
Posted: Sat Apr 28, 2007 9:21 am Post subject: |
|
|
hey there.
I got a problem with your patchset above 2.6.20-viper7: The Toshiba extras dont work any more.
Actually, the /proc/acpi/toshiba inferface is missing, so "fnfx" wont work.
Did you change something in viper8 and viper9 concerning acpi? I didnt notice anything in the changelog.
I haven't tried viper10 or 2.6.21-viper1, but I will soon.
Thanks
edit:
Code: | keinstein veit # dmesg | grep -i tosh
Toshiba System Managment Mode driver v1.11 26/9/2001
toshiba_acpi: Toshiba Laptop ACPI Extras version 0.19a-dev
toshiba_acpi: HCI method: \_SB_.VALD.GHCI
toshiba_acpi: failed to register misc device 181 ("toshiba") |
|
|
Back to top |
|
|
Pse Apprentice
Joined: 01 Mar 2005 Posts: 188 Location: by the plate river
|
Posted: Sat Apr 28, 2007 1:29 pm Post subject: |
|
|
darklegion wrote: | vipernicus wrote: | darklegion wrote: | This release seems to be using the bugged reiser4 patch as /bin/true was corrupted after a few hours of use.Use the reiser4 patch from viper4 instead.
This problem probably occurs with 2.6.21-nicus1 as well as it using the same patch. |
There is no nonbugged reiser4 patch, the fix for the 0'ed files hasn't been released by reiser4 developers yet. |
Are you sure? Its never happened with me on viper4 or dark6 (which I patched in the old viper4 reiser4 patch) and thats after weeks (months?) of use.It has ,however, happened with viper7 and it happened almost immediatly with viper10.Maybe its just a coincidence, though.I will note that booting back into a kernel with the older patch does not reverse the problem, but it doesn't seem to re-occur after emerging coreutils.
I'm currently running 2.6.21-nicus1 with the viper4 patch, with no problems yet.I'll stand corrected if it happens again, though |
Given that most find the R4 patch from Viper4 fairly stable and much better than the later versions, I guess it wouldn't have been too crazy to put it in in this release. Ah, well, too late... I'll be trying ext4 the following days, what's that 'dealloc' option I keep reading about? I can't seem to find any info =( Is it safe to use? |
|
Back to top |
|
|
buddabrod Apprentice
Joined: 15 Oct 2006 Posts: 241 Location: Germany
|
|
Back to top |
|
|
Pse Apprentice
Joined: 01 Mar 2005 Posts: 188 Location: by the plate river
|
Posted: Sat Apr 28, 2007 2:33 pm Post subject: |
|
|
buddabrod wrote: | http://en.wikipedia.org/wiki/Delayed_allocation
I tried ext4 and it's now my root fs. It's incredibly fast and more stable than reiser4 |
Thanks
So, the recommended options are 'extents,dealloc'. Should I add anything else for a root filesystem? 'noatime' perhaps? Any there any particular journal modes? My google-fu is considerately weak today
[EDIT] Nevermind, found them
Überspeed, less safe: 'noatime,extents,dealloc,data=writeback,nobh'
Cool: 'extents,dealloc'
Conservative: 'data=journal'
[EDIT-2] It's 'delalloc'. Thanks buddabrod!
Last edited by Pse on Sat Apr 28, 2007 4:16 pm; edited 1 time in total |
|
Back to top |
|
|
_sys/sid Guru
Joined: 27 Aug 2004 Posts: 346 Location: Asola (Mantova)
|
|
Back to top |
|
|
|