Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
OFFICIAL Zen Kernel Discussion - zen-kernel.org
View unanswered posts
View posts from last 24 hours

Goto page Previous  1, 2, 3 ... , 9, 10, 11  Next  
This topic is locked: you cannot edit posts or make replies.    Gentoo Forums Forum Index Unsupported Software
View previous topic :: View next topic  
Author Message
cheater1034
Veteran
Veteran


Joined: 09 Sep 2004
Posts: 1558

PostPosted: Sat Nov 14, 2009 8:17 pm    Post subject: Reply with quote

RealNC wrote:
Waninkoko wrote:
FatELF is not really dead.


Well, according to Ryan Gordon himself:

http://icculus.org/cgi-bin/finger/finger.pl?user=icculus&date=2009-11-03&time=19-08-04

"I think I'll declare FatELF done for now. I'll leave the project page up, but I imagine it'll only be for archeological purposes."

it looks grim. Anyway, even if it was alive and kicking, which it isn't, why put it in Zen?


We are totally aware (we know about the projects we put in zen ;))
We've talked to the maintainer and it will be maintained and we have it in zen.
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land.
Back to top
View user's profile Send private message
RealNC
Tux's lil' helper
Tux's lil' helper


Joined: 13 Apr 2008
Posts: 148

PostPosted: Sun Nov 15, 2009 6:35 am    Post subject: Reply with quote

The only thing sticking out from that post is that Gordon is a rude asshole, not much better than some kernel devs he was complaining about:

"There's nothing worse than hearing your name in the same sentence as Con Kolivas"

But let me ask, again, what does FatELF contribute to Zen exactly from a "Desktop" point of view?
Back to top
View user's profile Send private message
NotQuiteSane
Guru
Guru


Joined: 30 Jan 2005
Posts: 481
Location: Klamath Falls, Jefferson, USA, North America, Midgarth

PostPosted: Wed Nov 18, 2009 10:26 pm    Post subject: Reply with quote

cheater1034 wrote:
It matters with BFS, because it keeps the CPUs busy and working, which is not the case with CFS.

On BFS+quad core, if you specify -j1 you are giving 1 core the work, -j2 2 cores, -j4 would occupy all of the cores for compilation, using more jobs than you have cores isn't a good idea for BFS (overloading the CPU)


What about with distcc is is just '-j$Number-of-total-cores'?

a quick count shows 15 cores currently up. had to shut down the 40 node p2 cluster. got too expensive.

NQS
_________________
These opinions are mine, mine I say! Piss off and get your own.

As I see it -- An irregular blog, Improved with new location

To delete French language packs from system use 'sudo rm -fr /'
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6108
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Wed Nov 18, 2009 10:33 pm    Post subject: Reply with quote

NotQuiteSane wrote:
cheater1034 wrote:
It matters with BFS, because it keeps the CPUs busy and working, which is not the case with CFS.

On BFS+quad core, if you specify -j1 you are giving 1 core the work, -j2 2 cores, -j4 would occupy all of the cores for compilation, using more jobs than you have cores isn't a good idea for BFS (overloading the CPU)


What about with distcc is is just '-j$Number-of-total-cores'?

a quick count shows 15 cores currently up. had to shut down the 40 node p2 cluster. got too expensive.

NQS


yeah, just like Con wrote - that should be fine

make sure that there's not too much latency over the lines or packages that generally compile pretty fast - otherwise it'll take longer due to the overhead than usual

how does BFS scale with these amounts of cores, btw ? (as good as CFS or better ? :wink: )
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
NotQuiteSane
Guru
Guru


Joined: 30 Jan 2005
Posts: 481
Location: Klamath Falls, Jefferson, USA, North America, Midgarth

PostPosted: Wed Nov 18, 2009 10:41 pm    Post subject: Reply with quote

kernelOfTruth wrote:
yeah, just like Con wrote - that should be fine

make sure that there's not too much latency over the lines or packages that generally compile pretty fast - otherwise it'll take longer due to the overhead than usual

how does BFS scale with these amounts of cores, btw ? (as good as CFS or better ? :wink: )


can't say. I've run no tests, but I replaced the p2 node with a dual p4 xeon, and speed are close to same. running zen on all with similar base configuration, but have been running the old 2x+1 rule.

other thing is I've currently got (1) 32 bit ppc and (6) sparc cores in the mix, so that may change things too. I'd like to add the ARM in my router, but it's running DD-WRT, and don't feel like makeing / hunting down packages for distcc and cross compile for it. my rule is every machine works.

NQS
_________________
These opinions are mine, mine I say! Piss off and get your own.

As I see it -- An irregular blog, Improved with new location

To delete French language packs from system use 'sudo rm -fr /'
Back to top
View user's profile Send private message
HorizonXP
n00b
n00b


Joined: 13 Dec 2008
Posts: 7

PostPosted: Thu Nov 19, 2009 5:58 pm    Post subject: Reply with quote

Ok, I do development work on my Core i7 machine.

Why do I want to use zen-kernel?
Back to top
View user's profile Send private message
ppurka
Advocate
Advocate


Joined: 26 Dec 2004
Posts: 3256

PostPosted: Thu Nov 19, 2009 6:01 pm    Post subject: Reply with quote

HorizonXP wrote:
Ok, I do development work on my Core i7 machine.

Why do I want to use zen-kernel?
Read through the website, read through this thread. If you can't find a reason, then don't use it. Simple. :roll:
_________________
emerge --quiet redefined | E17 vids: I, II | Now using kde5 | e is unstable :-/
Back to top
View user's profile Send private message
cheater1034
Veteran
Veteran


Joined: 09 Sep 2004
Posts: 1558

PostPosted: Thu Nov 19, 2009 9:34 pm    Post subject: Reply with quote

HorizonXP wrote:
Ok, I do development work on my Core i7 machine.

Why do I want to use zen-kernel?


http://zen-kernel.org/documentation

If I made commission from getting users I may try to sell it to you (but i don't), it may or may not benefit you, there is plenty of information you can read yourself and figure out if you want to use it or not.
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land.
Back to top
View user's profile Send private message
bollucks
l33t
l33t


Joined: 27 Oct 2004
Posts: 606

PostPosted: Fri Nov 20, 2009 9:38 am    Post subject: Reply with quote

NotQuiteSane wrote:
cheater1034 wrote:
It matters with BFS, because it keeps the CPUs busy and working, which is not the case with CFS.

On BFS+quad core, if you specify -j1 you are giving 1 core the work, -j2 2 cores, -j4 would occupy all of the cores for compilation, using more jobs than you have cores isn't a good idea for BFS (overloading the CPU)


What about with distcc is is just '-j$Number-of-total-cores'?

a quick count shows 15 cores currently up. had to shut down the 40 node p2 cluster. got too expensive.

NQS


BFS may be efficient at using every core on the local machine, but distcc is hopeless at sending out enough work for other machines, so you'll find you need a LOT more work than you have cores, and in practice it won't scale remotely well. It's not entirely distcc's fault either as you can only parallelise so many things while waiting for other things to return back. I suggest at least double the total number of cores based on my experience and I had less total cores than you. Beyond that it didn't seem to scale any further.
Back to top
View user's profile Send private message
jw5801
Apprentice
Apprentice


Joined: 12 Jun 2008
Posts: 251
Location: Melbourne, Australia

PostPosted: Fri Nov 20, 2009 11:14 am    Post subject: Reply with quote

I'm using 2.6.31-zen8 with the drm-new branch (installed using the ebuild). After attempting to resume from suspend today, I got a kernel panic and a message on the screen reading:
Code:
journal commit I/O error


This is the first time I've seen this, and I've successfully suspended and resumed many times with this kernel. I couldn't see anything much in the syslog, the last message there came as I suspended the machine and was:
Code:
[drm] Num pipes: 4


My kernel .config. Not sure what else would be useful. Any tips for things I can look out for and take note of in case it happens again?
Back to top
View user's profile Send private message
SpectreDragon
Tux's lil' helper
Tux's lil' helper


Joined: 22 Aug 2008
Posts: 92
Location: Germany

PostPosted: Sat Nov 21, 2009 9:02 pm    Post subject: Reply with quote

Since half a week I get regular kernel panics with zen-sources.

It seems that this is caused by using my wireless connection with 2.6.32_rc7-zen1, currently trying with 2.6.32_rc6-zen1 and no problems so far. Unfortunately no kernel message is logged when the panic occurs, is there anything else I can do to verify my guess about the cause?

(corrected wrong kernel version, i'm not back using 2.6.31 but 2.6.32_rc6 which seems to have no problems)
Back to top
View user's profile Send private message
NotQuiteSane
Guru
Guru


Joined: 30 Jan 2005
Posts: 481
Location: Klamath Falls, Jefferson, USA, North America, Midgarth

PostPosted: Mon Nov 23, 2009 4:32 am    Post subject: Reply with quote

bollucks wrote:
NotQuiteSane wrote:
cheater1034 wrote:
It matters with BFS, because it keeps the CPUs busy and working, which is not the case with CFS.

On BFS+quad core, if you specify -j1 you are giving 1 core the work, -j2 2 cores, -j4 would occupy all of the cores for compilation, using more jobs than you have cores isn't a good idea for BFS (overloading the CPU)


What about with distcc is is just '-j$Number-of-total-cores'?

a quick count shows 15 cores currently up. had to shut down the 40 node p2 cluster. got too expensive.

NQS


BFS may be efficient at using every core on the local machine, but distcc is hopeless at sending out enough work for other machines, so you'll find you need a LOT more work than you have cores, and in practice it won't scale remotely well. It's not entirely distcc's fault either as you can only parallelise so many things while waiting for other things to return back. I suggest at least double the total number of cores based on my experience and I had less total cores than you. Beyond that it didn't seem to scale any further.


well, i've been using portage 2.2, and the flag --jobs=6, but for some reason the netbook is loking up under that. experimenting with --jobs=3 on it

NQS
_________________
These opinions are mine, mine I say! Piss off and get your own.

As I see it -- An irregular blog, Improved with new location

To delete French language packs from system use 'sudo rm -fr /'
Back to top
View user's profile Send private message
bollucks
l33t
l33t


Joined: 27 Oct 2004
Posts: 606

PostPosted: Mon Nov 23, 2009 8:52 am    Post subject: Reply with quote

NotQuiteSane wrote:
bollucks wrote:
NotQuiteSane wrote:
cheater1034 wrote:
It matters with BFS, because it keeps the CPUs busy and working, which is not the case with CFS.

On BFS+quad core, if you specify -j1 you are giving 1 core the work, -j2 2 cores, -j4 would occupy all of the cores for compilation, using more jobs than you have cores isn't a good idea for BFS (overloading the CPU)


What about with distcc is is just '-j$Number-of-total-cores'?

a quick count shows 15 cores currently up. had to shut down the 40 node p2 cluster. got too expensive.

NQS


BFS may be efficient at using every core on the local machine, but distcc is hopeless at sending out enough work for other machines, so you'll find you need a LOT more work than you have cores, and in practice it won't scale remotely well. It's not entirely distcc's fault either as you can only parallelise so many things while waiting for other things to return back. I suggest at least double the total number of cores based on my experience and I had less total cores than you. Beyond that it didn't seem to scale any further.

well, i've been using portage 2.2, and the flag --jobs=6, but for some reason the netbook is loking up under that. experimenting with --jobs=3 on it


I was NOT recommending increasing the number of jobs to more than the number of CPUs for a local compile. I suggest sticking with the BFS guidelines of jobs==CPUs (or at most +1). I was only suggesting increasing the number of jobs with DISTCC because -that- is inefficient at distributing jobs evenly and keeping cores busy. It's also hopeless and keeping different speed cores fairly loaded and whole compiles tend to lag waiting for slower CPUs. Follow the guidelines carefully and order your DISTCC hosts in fastest to slowest order and possibly not even put jobs on the local machine that's serving the jobs out.
Back to top
View user's profile Send private message
cheater1034
Veteran
Veteran


Joined: 09 Sep 2004
Posts: 1558

PostPosted: Tue Nov 24, 2009 2:51 am    Post subject: Reply with quote

jw5801 wrote:
I'm using 2.6.31-zen8 with the drm-new branch (installed using the ebuild). After attempting to resume from suspend today, I got a kernel panic and a message on the screen reading:
Code:
journal commit I/O error


This is the first time I've seen this, and I've successfully suspended and resumed many times with this kernel. I couldn't see anything much in the syslog, the last message there came as I suspended the machine and was:
Code:
[drm] Num pipes: 4


My kernel .config. Not sure what else would be useful. Any tips for things I can look out for and take note of in case it happens again?


Maybe you have some filesystem corruption, i can't tell anything from that information though - the full panic would be more useful :O
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land.
Back to top
View user's profile Send private message
jw5801
Apprentice
Apprentice


Joined: 12 Jun 2008
Posts: 251
Location: Melbourne, Australia

PostPosted: Tue Nov 24, 2009 3:04 am    Post subject: Reply with quote

cheater1034 wrote:
jw5801 wrote:
I'm using 2.6.31-zen8 with the drm-new branch (installed using the ebuild). After attempting to resume from suspend today, I got a kernel panic and a message on the screen reading:
Code:
journal commit I/O error


This is the first time I've seen this, and I've successfully suspended and resumed many times with this kernel. I couldn't see anything much in the syslog, the last message there came as I suspended the machine and was:
Code:
[drm] Num pipes: 4


My kernel .config. Not sure what else would be useful. Any tips for things I can look out for and take note of in case it happens again?


Maybe you have some filesystem corruption, i can't tell anything from that information though - the full panic would be more useful :O


Well, haven't had it happen again. I'll run a full fsck when I get the chance to check for filesystem corruption.

I'm assuming it was a kernel panic as it started flashing the numlock and capslock lights as me, all the output I got was a black screen with the one line in the top left hand corner, nothing in the logs, so I'm not sure how to get at any more information.
Back to top
View user's profile Send private message
cheater1034
Veteran
Veteran


Joined: 09 Sep 2004
Posts: 1558

PostPosted: Tue Nov 24, 2009 3:30 am    Post subject: Reply with quote

jw5801 wrote:
cheater1034 wrote:
jw5801 wrote:
I'm using 2.6.31-zen8 with the drm-new branch (installed using the ebuild). After attempting to resume from suspend today, I got a kernel panic and a message on the screen reading:
Code:
journal commit I/O error


This is the first time I've seen this, and I've successfully suspended and resumed many times with this kernel. I couldn't see anything much in the syslog, the last message there came as I suspended the machine and was:
Code:
[drm] Num pipes: 4


My kernel .config. Not sure what else would be useful. Any tips for things I can look out for and take note of in case it happens again?


Maybe you have some filesystem corruption, i can't tell anything from that information though - the full panic would be more useful :O


Well, haven't had it happen again. I'll run a full fsck when I get the chance to check for filesystem corruption.

I'm assuming it was a kernel panic as it started flashing the numlock and capslock lights as me, all the output I got was a black screen with the one line in the top left hand corner, nothing in the logs, so I'm not sure how to get at any more information.


you're compiled with frame pointers and other useful things right?

if not i may beat you down.
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land.
Back to top
View user's profile Send private message
jw5801
Apprentice
Apprentice


Joined: 12 Jun 2008
Posts: 251
Location: Melbourne, Australia

PostPosted: Tue Nov 24, 2009 3:49 am    Post subject: Reply with quote

cheater1034 wrote:
you're compiled with frame pointers and other useful things right?

if not i may beat you down.


Um... I thought I was, but apparently not. I am now, though. :oops:

* Prepares for beat down *
Back to top
View user's profile Send private message
pappy_mcfae
Watchman
Watchman


Joined: 27 Dec 2007
Posts: 5997
Location: Pomona, California.

PostPosted: Tue Nov 24, 2009 8:49 am    Post subject: Reply with quote

I've just finished and uploaded seeds for zen source kernels. Stop on down and take a look. See my sig for a link to the site.

Blessed be!
Pappy
_________________
This space left intentionally blank, except for these ASCII symbols.
Back to top
View user's profile Send private message
hallabro
n00b
n00b


Joined: 12 Oct 2008
Posts: 24

PostPosted: Tue Nov 24, 2009 12:19 pm    Post subject: Reply with quote

pappy_mcfae wrote:
I've just finished and uploaded seeds for zen source kernels. Stop on down and take a look. See my sig for a link to the site.

Blessed be!
Pappy


Thanks for the seeds, great work!

Although I must ask, how come rr_interval is set to 1? Isn't the default value quite a safe value (CONFIG_BFS_RR_INTERVAL=1)?

Keep it up. :)
Back to top
View user's profile Send private message
pappy_mcfae
Watchman
Watchman


Joined: 27 Dec 2007
Posts: 5997
Location: Pomona, California.

PostPosted: Tue Nov 24, 2009 5:52 pm    Post subject: Reply with quote

I like fast. The settings are actually in use on my machines. I've yet to have issues with those settings.

BB!
P
_________________
This space left intentionally blank, except for these ASCII symbols.
Back to top
View user's profile Send private message
cheater1034
Veteran
Veteran


Joined: 09 Sep 2004
Posts: 1558

PostPosted: Tue Nov 24, 2009 8:10 pm    Post subject: Reply with quote

jw5801 wrote:
cheater1034 wrote:
you're compiled with frame pointers and other useful things right?

if not i may beat you down.


Um... I thought I was, but apparently not. I am now, though. :oops:

* Prepares for beat down *


Yes do that, and then look for any useful information perhaps in dmesg, etc. (i don't expect a kernel bug tho)

@pappy
Thanks!
default rr_interval is 6ms (the minimum human processing speed)
I'm not sure if I would jump down to 1ms - 3ms minimum - but in general the default value is the best option.
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land.
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6108
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Tue Nov 24, 2009 11:40 pm    Post subject: Reply with quote

cheater1034 wrote:
jw5801 wrote:
cheater1034 wrote:
you're compiled with frame pointers and other useful things right?

if not i may beat you down.


Um... I thought I was, but apparently not. I am now, though. :oops:

* Prepares for beat down *


Yes do that, and then look for any useful information perhaps in dmesg, etc. (i don't expect a kernel bug tho)

@pappy
Thanks!
default rr_interval is 6ms (the minimum human processing speed)
I'm not sure if I would jump down to 1ms - 3ms minimum - but in general the default value is the best option.


according to the documentation a lower value might gain even better interactivity (which I noticed today under heavy multi-tasking load) but it might also lower the maximum throughput as a downside

so take your pick - it's not hardcoded so you can change it manually later anyway
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
cheater1034
Veteran
Veteran


Joined: 09 Sep 2004
Posts: 1558

PostPosted: Wed Nov 25, 2009 2:04 am    Post subject: Reply with quote

kernelOfTruth wrote:
cheater1034 wrote:
jw5801 wrote:
cheater1034 wrote:
you're compiled with frame pointers and other useful things right?

if not i may beat you down.


Um... I thought I was, but apparently not. I am now, though. :oops:

* Prepares for beat down *


Yes do that, and then look for any useful information perhaps in dmesg, etc. (i don't expect a kernel bug tho)

@pappy
Thanks!
default rr_interval is 6ms (the minimum human processing speed)
I'm not sure if I would jump down to 1ms - 3ms minimum - but in general the default value is the best option.


according to the documentation a lower value might gain even better interactivity (which I noticed today under heavy multi-tasking load) but it might also lower the maximum throughput as a downside

so take your pick - it's not hardcoded so you can change it manually later anyway


I'm aware of that - 6ms is the ideal value, 3ms is what I would recommend for better interactivity personally.
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land.
Back to top
View user's profile Send private message
pappy_mcfae
Watchman
Watchman


Joined: 27 Dec 2007
Posts: 5997
Location: Pomona, California.

PostPosted: Wed Nov 25, 2009 4:18 am    Post subject: Reply with quote

Quote:
I'm not sure if I would jump down to 1ms - 3ms minimum - but in general the default value is the best option.


Oh, I don't know. That setting has been working rather well for me. I always crank to the max, and then back down if needed. So far, it's not been needed.

It makes a difference in how well this machine runs openGL screen savers. It really does improve snappiness. I haven't seen anything bad coming from it since it became a zen option.

Blessed be!
Pappy
_________________
This space left intentionally blank, except for these ASCII symbols.
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6108
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Fri Nov 27, 2009 1:01 pm    Post subject: Reply with quote

cheater1034 wrote:


I'm aware of that - 6ms is the ideal value, 3ms is what I would recommend for better interactivity personally.


I second that recommendation !


for those folks who have a laggy mouse-pointer under load / no load with X and BFS:

try compiling your kernel with optimized flags this (seemingly) made the problem go away for me

--> that was the main annoying difference / regression between 2.6.31-zen8 and 2.6.31-zen9 that was noticable for me:

a laggy mouse pointer ! (one of the little changes was reverting of the custom flags patch and re-applying that "fixed" that for me)


apply the custom flags patch mentioned some posts above and try to use some performance flags:

e.g.
Quote:
-O2 -march=native -pipe -fomit-frame-pointer --param max-gcse-passes=8 -fno-strict-overflow -fno-delete-null-pointer-checks -fno-tree-vrp -mno-align-stringops -minline-stringops-dynamically -mno-push-args -fno-ident


green flags = hardened, safe / for sanity's sake

indigo flags = x86_64 safe flags for better optimization, size savings, etc.

the others should be safe too


I have much more aggressive flags (that are heavily tested but won't post them) - so if those posted flags don't fix it for you try to "rice" some more if you dare whether that fixes it or come up with a fix :wink:
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
Display posts from previous:   
This topic is locked: you cannot edit posts or make replies.    Gentoo Forums Forum Index Unsupported Software All times are GMT
Goto page Previous  1, 2, 3 ... , 9, 10, 11  Next
Page 10 of 11

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum