2.6.11-rc1-nitro0: Back To The Basics

mirko_3 · l33t Joined: 02 Nov 2003 Posts: 605 Location: Birreria

I thought that 4k stacks was supposed to make memory management easier; yet now I see 16k stacks... why would someone want that?? Could someone please explain?
Thank you
_________________
Non fa male! Non fa male!

chunderbunny · Posted: Mon Jan 24, 2005 2:28 pm Post subject:

As far as I understand it the stack size is trade off of memory efficincy vs speed. In very general terms, smaller stacks use memory more efficiently, whereas larger sticks use more memory but should give better performance.

discomfitor · l33t Joined: 21 Feb 2003 Posts: 927 Location: None

Sorry about the delay between releases; there seems to have been a break in communication regarding the latest release. It will be out as soon as possible.

List any patches that you want included here. So far I've noted the xbox patches (most likely going to be optional with a use flag) and LIRC. And there's a new stack patch coming which should entertain everyone.
_________________
There is no substitute for experience.
Imperfection indicates a lack of effort.

pifactorial · Posted: Mon Jan 24, 2005 7:26 pm Post subject: force feedback

It's probably not worth the effort, but I've been longing for a kernel integrating this patch:
http://madfab.free.fr/ff/

I say it's probably not worth the effort, because the patch was written for kernel 2.4.19, and I'm probably one of the few nitro users with said joystick.

Nonetheless, I'd love you forever if you did it...

thx

discomfitor · l33t Joined: 21 Feb 2003 Posts: 927 Location: None

pifactorial: I'll do my best

_________________
There is no substitute for experience.
Imperfection indicates a lack of effort.

rmh3093 · Posted: Mon Jan 24, 2005 7:43 pm Post subject:

well if you wanted to be nice you could include these patches for me

http://www.personal.uni-jena.de/~p1stja/linux/cpad-old/cpad-kpatch.html

and

http://memebeam.org/free-software/toshiba_acpi/toshiba_acpi_0.18-dev_toshiba_test4.patch
_________________
Do not meddle in the affairs of wizards, for they are subtle and quick to anger.

pifactorial · Posted: Mon Jan 24, 2005 8:57 pm Post subject: Kudos

We really work you too hard, Darckness! Just wanted to give you some general kudos for keeping such a great attitude, and making such a great kernel!

BTW... do you have a real job?

Keep up the good work :wink:

fimbulvetr · n00b Joined: 03 Feb 2004 Posts: 51 Location: South Dakota

I actually stopped using nitro (or upgrading, rather) because all of a sudden people didn't want to include the good patches anymore (like swsup) and because they decided they were going to only release one a month or something. Are you telling me the *REAL* nitro is back? The one that makes me feel like I'm living dangerously? You're doing RCs again?
I use(d) nitro because:
a. It's bleeding edge, I can help test stuff.
b. It's got some good patches.
c. People put a lot of work into it.
Nitro for the past two months has eliminated some of that.

Edit:
I'm not trying to 'dis anyone, nor piss them off, I just miss the way it was.
_________________
:wq

discomfitor · l33t Joined: 21 Feb 2003 Posts: 927 Location: None

rmh3093: I'll give them a try

pifactorial: No, I don't have a real job.

fimbulvetr: Perhaps you've missed the few nitro releases; they've made ME feel like I was living dangerously, and I'm the kind of guy who plays filesystem Russian Roulette for fun. Of course, there's also the more stable versions so that people who like to live life a good distance away from the edge can do so. If you want a patch added, tell me. That's the only way I'll know if it's worth adding/it exists.
_________________
There is no substitute for experience.
Imperfection indicates a lack of effort.

Xithix · Apprentice Joined: 31 Dec 2004 Posts: 228

Which nitro versions are more stable and which are more bleeding edge?

Is it merely the "want-experimental" USE flag?

rmh3093 · Posted: Mon Jan 24, 2005 10:24 pm Post subject:

hey idk about u guys, but I think we can ditch staircase cpu scheduler and use the
PlugSched from https://sourceforge.net/projects/cpuse/

it adds a menu to general setup to choose cpu scheduler, im used the zaphod now it runs nice
i have it applied to the default 2.6.10 with genetic lib and genetic-io and genetic-as and some random ck patches

so far so good
_________________
Do not meddle in the affairs of wizards, for they are subtle and quick to anger.

discomfitor · l33t Joined: 21 Feb 2003 Posts: 927 Location: None

Xithix: For experimental use "want-experimental". Default is "stable".

rmh3093: We can't "ditch" staircase. That's one of the founding points of nitro-sources. However you are correct in saying that zaphod should be integrated as well. For that reason I've looked into a few options. The one that I will (most likely) be using for this next release is another use flag which will optionally remove staircase and insert (genetic) zaphod instead. There will be a better alternative to it in the future

As for progress on the next nitro, all I'm waiting for is the ck version for rc2. I tried to hack the newest ck patches for it, but it's nothing that a simple hack job can do. Most of the code has been completely redone, meaning that there's nothing I can do. Rest assured though that as soon as new ck stuff comes out, so will nitro.

freelight: If you're going to try that, use 2.6.10-nitro4; That's probably the most stable kernel I've ever made/used.
_________________
There is no substitute for experience.
Imperfection indicates a lack of effort.

freelight · Posted: Tue Jan 25, 2005 12:05 am Post subject:

Just a little question. I use nitro on my dekstop and laptop and usually don't have problems with it, but how insane would it be of me to put it on a (small) production server?

Edit: Thanks for the advice, I'll put 2.6.10-nitro4 on it now and see how it works, then I'll try -dark4 later and compare. (The hardware isn't exactly new, and I do want to get the most out of it.)

predatorfreak · Posted: Tue Jan 25, 2005 12:50 am Post subject:

rmh3093 · Posted: Tue Jan 25, 2005 1:57 am Post subject:

well that plugin scheduler gives u a nice menu
CONFIG_PLUGSCHED=y
# CONFIG_CPUSCHED_DEFAULT_INGO is not set
# CONFIG_CPUSCHED_DEFAULT_STAIRCASE is not set
# CONFIG_CPUSCHED_DEFAULT_SPA_NF is not set
CONFIG_CPUSCHED_DEFAULT_ZAPHOD=y

i dont know if that staircase is different than ck's modified but so far i like how zaphod is behaving... i was gonna try unixbench and try some benchmarks but i dont know the flavor unx I use, i guess sysv and it didnt get a result from any test so i will try it out with bsd again

EDIT: you dont have to use "USE" flags because it there is a config option the kernel
_________________
Do not meddle in the affairs of wizards, for they are subtle and quick to anger.

Xithix · Apprentice Joined: 31 Dec 2004 Posts: 228

So what difference does it make which scheduler you use? Can you choose one or another, assuming the kernel includes both, and use that one exclusively?

discomfitor · l33t Joined: 21 Feb 2003 Posts: 927 Location: None

rmh3093: The staircase included with plugsched is different than the one from ck (much older). Until it's updated I probably won't be using it.
_________________
There is no substitute for experience.
Imperfection indicates a lack of effort.

bollucks · l33t Joined: 27 Oct 2004 Posts: 606

rmh3093 · Posted: Tue Jan 25, 2005 3:34 am Post subject:

bollucks · l33t Joined: 27 Oct 2004 Posts: 606

discomfitor · l33t Joined: 21 Feb 2003 Posts: 927 Location: None

Hrmm...I haven't checked it in a few days; guess a new release came out! Shouldn't be too hard to cram it in there...
_________________
There is no substitute for experience.
Imperfection indicates a lack of effort.

fallow · Posted: Tue Jan 25, 2005 8:11 am Post subject:

imho staircase is better for destkop use than genetic zaphod .
genetic zaphod is better for server use . ( yes I know Bollucks , staircase is also good for servers , compute etc.

) , but zaphod with all his options is better configurable at runtime and can be better tuned by genetic algoritm.

but - J.Moilanen is working on genetic lib optimizations for desktop , now is problems with interactivity : it`s ok when child gotten mutated , find better fit for workload. Child might have
picked bad settings for latency, and system slow
down for a couple of seconds. Settings will return to the good , but latency was been bad at the moment.

that is now main reason , J.Moilanen is working on a couple
ways to fix this problem.

Now staircase is better

- low latency
My friends are trying gen zaphod gl3 based with small lan servers now and will do some benchmarsk to previous scheds.

cheers.
_________________
"Time is a companion that goes with us on a journey. It reminds us to cherish each moment, because it will never come again. What we leave behind is not as important as how we have lived" J-L. Picard

bollucks · l33t Joined: 27 Oct 2004 Posts: 606

predatorfreak · Posted: Tue Jan 25, 2005 2:58 pm Post subject:

Well we're talking about scheduler's, I'v had some ideas of modifing genetic for desktop use and modifing staircase to utilise my modified genetic lib. Though it would likely take me a while to complete the idea, i think it would be worth investing some time in.

~predatorfreak
_________________
System: predatorbox
Distro: Arch Linux x86_64
Current projects: blackhole, convmedia and anything else I cook up.