| View previous topic :: View next topic |
| Author |
Message |
Tiger683
Veteran
Joined: 08 Jan 2005
Posts: 1347
Location: Heffner's House
|
 Posted: Sun Jul 17, 2005 11:40 am Post subject: |
 |
|
only thing to emerge -C is the "media-sound/alsa-driver", other alsa stuff should be updated (if not already up2date) to 1.0.9b (headers and libs in any case),
when still not working, compile the kernel driver as builtin, not as module.
T
_________________
Retired gentoo user |
|
| Back to top |
|
 |
chronophobic
Apprentice
Joined: 07 Mar 2005
Posts: 237
Location: Sofia, Bulgaria
|
| Posted: Sun Jul 17, 2005 1:59 pm Post subject: |
|
|
| Chronophobic wrote: | ... And it still doesn't work for some reason  . I'll try to investigate and get back to you ... |
Just one word... stoopid (I know it's not in the dictionary, but it fits the situation). I should configure my kernel more carefully next time. All works top notch now, I'll go on to play with the staircase scheduler (I don't expect mind-numbing performance surge on a 3-year old laptop tho)
_________________
Confutatis maledictis, flammis acribus addictis!
Last edited by chronophobic on Tue Jul 19, 2005 12:08 am; edited 2 times in total |
|
| Back to top |
|
|
Kensai
Guru
Joined: 09 Mar 2004
Posts: 569
Location: Puerto Rico
|
| Posted: Sun Jul 17, 2005 2:20 pm Post subject: |
|
|
In the staircase cpu-scheduler
all the values in 1 means better performance?
_________________
Gentoo: Gigabyte: nFORCE 2: nVIDIA GeForce 6600: AMD Athlon XP 3200+
Leaving the above specs to immortalize the first system I Installed Gentoo on! |
|
| Back to top |
|
|
Tiger683
Veteran
Joined: 08 Jan 2005
Posts: 1347
Location: Heffner's House
|
| Posted: Sun Jul 17, 2005 3:06 pm Post subject: Re: 2.6.12-nitro5 "Make my day" |
|
|
| Kensai wrote: | In the staircase cpu-scheduler
all the values in 1 means better performance? |
ugh! 
look here:
| Tiger683 wrote: |
settings are in these ranges:
timeslice = {1 to 20}
timeslice_factor = {10 to 26}
burst_factor = {1 to 10}
Can also be changed on the fly through /proc/sys/kernel/{timeslice,timeslice_factor,burst_factor} with an echo command too
|
so twice no!
1) no, because the settigs for timeslice_factor smaller than 10 are discarded.
2) talking about "performance" here is ricy and wrong, i would rather say responsivity aka. interactivity.
This also implies longer times for non-interactive tasks to get finished, so the _"performance"_ for a big computing server
will actually be worse.....
And again people:
READ THE FSCKING RELEASE NOTES __CAREFULLY__ !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
This patchset is about freedom, so i give you the freedom to mess with adaptable settings of the scheduler,
but also the possibility to f*ck up your desktop, so be carefull......
no hard feelings, but most of you will want to leave the defaults or decrease (/increase) settings slowly via /proc to
find best settings for THEIR average load first, and then recompile with these as default.....
but then again: this is no magic, and wont make your desktop fly !!!!!!!!! (the note about initial patch for nitro3 was partly a little joke  , this is just fine-tuning... )
_________________
Retired gentoo user |
|
| Back to top |
|
|
jannis
Guru
Joined: 05 Dec 2004
Posts: 340
Location: Germany / Bavaria / Aschaffenburg
|
| Posted: Sun Jul 17, 2005 4:10 pm Post subject: |
|
|
I will just post my experiences here:
1 - 10 - 1 ain't killing or freezing my system but it ain't better than without the timeslice-patch. actually I have more sound buffer underruns than without it.
3 - 12 - 3 seems pretty good for my sys (AMD XP 2100+; 512 MB DDR RAM)
I have less fps in quake3 but my system is more responsive so I can play even better. (with mouse-polling @ 2) |
|
| Back to top |
|
|
Kensai
Guru
Joined: 09 Mar 2004
Posts: 569
Location: Puerto Rico
|
| Posted: Sun Jul 17, 2005 5:02 pm Post subject: |
|
|
HEHEHE sorry wrong thoughts ran trhough my head the best performance setting I was using was. 5 - 10 - 1. I now changed to 3 - 10 - 1 to see. Thanks tiger683.
_________________
Gentoo: Gigabyte: nFORCE 2: nVIDIA GeForce 6600: AMD Athlon XP 3200+
Leaving the above specs to immortalize the first system I Installed Gentoo on! |
|
| Back to top |
|
|
m0p
Apprentice
Joined: 20 Jun 2005
Posts: 205
Location: en_GB
|
| Posted: Sun Jul 17, 2005 5:28 pm Post subject: |
|
|
| What would be the best setup for a P4 3.6GHz system with 2GB of RAM? |
|
| Back to top |
|
|
jannis
Guru
Joined: 05 Dec 2004
Posts: 340
Location: Germany / Bavaria / Aschaffenburg
|
| Posted: Sun Jul 17, 2005 6:06 pm Post subject: |
|
|
| m0p just try it out and play around. echo some values to the /proc/.... files and observe your PC how well it performs. |
|
| Back to top |
|
|
m0p
Apprentice
Joined: 20 Jun 2005
Posts: 205
Location: en_GB
|
| Posted: Sun Jul 17, 2005 6:09 pm Post subject: |
|
|
I really don't feel like experimenting. If anyone has tried it with a simaler setup to mine, and had good results, then I'll try those.
EDIT: Just a question, how do I enable the staircase scheduler? Usually I use the Genetic one, is staircase any better? |
|
| Back to top |
|
|
Tiger683
Veteran
Joined: 08 Jan 2005
Posts: 1347
Location: Heffner's House
|
| Posted: Sun Jul 17, 2005 7:03 pm Post subject: |
|
|
m0p: you have largely misunderstood and thus mixed up the concepts.
for the first, id suggest you just leave the defaults in that case.....
But just to provide a clarification:
genetic anticipatory scheduler is an io-scheduler, which governs the access to your io-devices(hdd, input devices, other peripherals),
but most noticeable effect is on the performance of block-devices within certain load-patterns, whereas
staircase is the cpu scheduler which governs the cpu-time which the tasks on your machine get dependent on priority etc.
So, basically these are two COMPLETELY different things.
This kernel has choice between 5 io-schedulers and 1 cpu cheduler:
io-schedulers:
==============
anticipatory without genetic
anticipatory with genetic
cfq
deadline
noop
cpu-scheduler:
==============
staircase
.
=> staircase scheduler is always selected and you cannot choose anything else nor deselect it.
cheers
T
_________________
Retired gentoo user |
|
| Back to top |
|
|
rjmalagon
n00b
Joined: 16 Jul 2005
Posts: 7
Location: Cuernavaca, Mexico
|
| Posted: Sun Jul 17, 2005 7:59 pm Post subject: |
|
|
The problem is the sin of too much ram (ATI-DRIVERS doesn't like 512MB+)
2GB Without DRI or 512MB with DRI... thanks ATI!
Waiting next drivers... |
|
| Back to top |
|
|
jannis
Guru
Joined: 05 Dec 2004
Posts: 340
Location: Germany / Bavaria / Aschaffenburg
|
| Posted: Sun Jul 17, 2005 8:21 pm Post subject: |
|
|
| What's the problem with 512MB+ and the ATI-Drivers? Ok, I don't have to use them, my Radeon 9000 Pro works great with the builtin/xorg DRI-Drivers. Is HIGHMEM a problem for the drivers? |
|
| Back to top |
|
|
borkdox
Tux's lil' helper
Joined: 16 Jan 2004
Posts: 123
|
| Posted: Sun Jul 17, 2005 10:51 pm Post subject: The Return of The Sound Skip |
|
|
does anyone experiences sound skips while using reiser4 and doing a high io operation like copying a 4gb+ folder from a partition in another harddrive, to the reiser4(/) partition?
Sound files reside in the same partition the copying is being done into. I am using nitro's default staircase parameters, altough this operation is not a highly CPU consuming one, im saying that just in case. i am using cfq io sched, tried the others but soudn still skips...
Is there any fix for this? Is anyone else experiencing this?
Sounds doesn't skip while doing a compile(if it does is very light, or unoticeable), or doign emerge sync. I am using rhythmbox as my music player.
Otherwise i have been amazed with reiser4 performance, Gnome (portage's gnome-light made my day) loads up in 8 seconds and my machine goes from Grub to GDM in about 50 seconds (with wireless networking and lot other services). |
|
| Back to top |
|
|
Tiger683
Veteran
Joined: 08 Jan 2005
Posts: 1347
Location: Heffner's House
|
| Posted: Mon Jul 18, 2005 6:38 am Post subject: |
|
|
the cause is reiser4, this fs is a cpu hog like none other, noticeable on some not-so-performant cpus....
solution: ext3 + dir_index (B*-trees)
look at this thread for howto:
https://forums.gentoo.org/viewtopic-t-305871.html
cheers
T
_________________
Retired gentoo user |
|
| Back to top |
|
|
martin.k
Guru
Joined: 28 Nov 2004
Posts: 493
Location: Wylatowo, Polska
|
| Posted: Mon Jul 18, 2005 8:31 am Post subject: |
|
|
| Tiger683 wrote: | m0p: you have largely misunderstood and thus mixed up the concepts.
for the first, id suggest you just leave the defaults in that case.....
io-schedulers:
==============
anticipatory without genetic
anticipatory with genetic
cfq
deadline
noop
cpu-scheduler:
==============
staircase
=> staircase scheduler is always selected and you cannot choose anything else nor deselect it.
cheers
T |
I think that fallow did some genetics for nicksched CPU Scheduler from Nick Piggin  (2.6.12-rc1 or sthg)
You can always use PlugSched from P.Williams and select IO and CPU scheduler via kernel BOOTPARAMS
But you know that all
_________________
linux-2.6.17 +ck +R4 +lockless +genetic-as +... więcej nie pamiętam
LRU #299256 |
|
| Back to top |
|
|
Tiger683
Veteran
Joined: 08 Jan 2005
Posts: 1347
Location: Heffner's House
|
| Posted: Mon Jul 18, 2005 8:37 am Post subject: |
|
|
| martin.k wrote: | | Tiger683 wrote: | m0p: you have largely misunderstood and thus mixed up the concepts.
for the first, id suggest you just leave the defaults in that case.....
io-schedulers:
==============
anticipatory without genetic
anticipatory with genetic
cfq
deadline
noop
cpu-scheduler:
==============
staircase
=> staircase scheduler is always selected and you cannot choose anything else nor deselect it.
cheers
T |
I think that fallow did some genetics for nicksched CPU Scheduler from Nick Piggin (2.6.12-rc1 or sthg)
You can always use PlugSched from P.Williams and select IO and CPU scheduler via kernel BOOTPARAMS
But you know that all |
Haven't seen plugsched in the patchlist  
as a sidenote: genetic algorithms in a cpu scheduler are useless at best
_________________
Retired gentoo user |
|
| Back to top |
|
|
martin.k
Guru
Joined: 28 Nov 2004
Posts: 493
Location: Wylatowo, Polska
|
| Posted: Mon Jul 18, 2005 8:50 am Post subject: |
|
|
| Tiger683 wrote: |
Haven't seen plugsched in the patchlist
as a sidenote: genetic algorithms in a cpu scheduler are useless at best |
Sorry  This is 2.6.12-nitro5 "Make my day" node and plugsched isn't in the patchlist.
I should have been more precise:
| Quote: |
m0p: There's a patch form P.Williams PlugSched that lets you use
different cpu schedulers, but it's not in the nitro5 patchlist
So you can only use staircase cpu scheduler |
_________________
linux-2.6.17 +ck +R4 +lockless +genetic-as +... więcej nie pamiętam
LRU #299256 |
|
| Back to top |
|
|
r0bertz
Retired Dev
Joined: 13 Sep 2004
Posts: 257
Location: HKSAR,China
|
| Posted: Mon Jul 18, 2005 2:17 pm Post subject: |
|
|
Nitro5 is great!!!
I switch to it
Thx to Tiger!
Btw, my integral SD card reader only works after applying this
http://projects.drzeus.cx/wbsd/sd.php
 |
|
| Back to top |
|
|
brot
Guru
Joined: 06 Apr 2004
Posts: 322
|
| Posted: Mon Jul 18, 2005 3:54 pm Post subject: ... |
|
|
I just wanted to say that -nitro5 works great here.
(with nvidia,vmware,captive-ntfs,..)
Thanks for releasing such a nice kernel ( and for releasing it so fast )
brot |
|
| Back to top |
|
|
Kensai
Guru
Joined: 09 Mar 2004
Posts: 569
Location: Puerto Rico
|
| Posted: Mon Jul 18, 2005 4:22 pm Post subject: |
|
|
it works great here but my FPS aren't stable anymore.
_________________
Gentoo: Gigabyte: nFORCE 2: nVIDIA GeForce 6600: AMD Athlon XP 3200+
Leaving the above specs to immortalize the first system I Installed Gentoo on! |
|
| Back to top |
|
|
jannis
Guru
Joined: 05 Dec 2004
Posts: 340
Location: Germany / Bavaria / Aschaffenburg
|
| Posted: Mon Jul 18, 2005 5:06 pm Post subject: |
|
|
Well Kensai as I said before my FPS are a bit lower than before but the system reacts faster and that's worth the less FPS.
nitro5 rules ^^ |
|
| Back to top |
|
|
m0p
Apprentice
Joined: 20 Jun 2005
Posts: 205
Location: en_GB
|
| Posted: Mon Jul 18, 2005 5:47 pm Post subject: |
|
|
| Tiger683 wrote: | m0p: you have largely misunderstood and thus mixed up the concepts.
for the first, id suggest you just leave the defaults in that case.....
But just to provide a clarification:
genetic anticipatory scheduler is an io-scheduler, which governs the access to your io-devices(hdd, input devices, other peripherals),
but most noticeable effect is on the performance of block-devices within certain load-patterns, whereas
staircase is the cpu scheduler which governs the cpu-time which the tasks on your machine get dependent on priority etc.
So, basically these are two COMPLETELY different things.
This kernel has choice between 5 io-schedulers and 1 cpu cheduler:
io-schedulers:
==============
anticipatory without genetic
anticipatory with genetic
cfq
deadline
noop
cpu-scheduler:
==============
staircase
.
=> staircase scheduler is always selected and you cannot choose anything else nor deselect it.
cheers
T |
Thanks for clearing up. So, is there a better scheduler than staircase? Or is that pretty much the best availible? |
|
| Back to top |
|
|
borkdox
Tux's lil' helper
Joined: 16 Jan 2004
Posts: 123
|
| Posted: Mon Jul 18, 2005 6:27 pm Post subject: |
|
|
| Tiger683 wrote: | the cause is reiser4, this fs is a cpu hog like none other, noticeable on some not-so-performant cpus....
solution: ext3 + dir_index (B*-trees)
look at this thread for howto:
https://forums.gentoo.org/viewtopic-t-305871.html
cheers
T |
How can reiser4 be a CPU hog in an Athlon 64 3000+@3500+(running 64bit), 1gig of ram??
Or is it??
I ve always used reiserfs, wich has decent IO and decent interactivity, but after seen fallow's recommendation(ext3+dir_index) decided to try that and was not happy with the degradation in loadtimes/bootimes/etc.. so decide to try reiser4, and was amazed with the IO performance, but not with the interactivity... |
|
| Back to top |
|
|
m0p
Apprentice
Joined: 20 Jun 2005
Posts: 205
Location: en_GB
|
| Posted: Mon Jul 18, 2005 7:17 pm Post subject: |
|
|
| ext3 ISN'T slow. Especially with tweaks. ReiserFS may be faster, but it's incredibly unstable. Try pulling the plug on a Reiser4 system while doing an intensive fs activity. I gaurantee you'll lose alot of data. And, with ReiserFS 3, you would lose data randomly, like having bad sectors, but they are on the fs, rather than physical. I can randomly lose power to my ext3 filesystems and not lose any data. The most loss I have ever had has been 3 orphaned inodes, which was in a fs intensive activity. |
|
| Back to top |
|
|
whitesouls
Guru
Joined: 19 Nov 2004
Posts: 358
Location: In Front of My Laptop
|
| Posted: Mon Jul 18, 2005 7:22 pm Post subject: |
|
|
 Yo 'T' you have my deepest respect man..thanks for helping me out with the problem... Now my laptop is breathing nitro .
To M0P, any articles or evidence to support your statement since your saying that ReiserFS is unstable
| Quote: | | ReiserFS may be faster, but it's incredibly unstable. |
_________________
whitesouls
Please insert the [SOLVED] tag if your problem is solved in your respective thread. |
|
| Back to top |
|
|
|
Unsupported Software
