| View previous topic :: View next topic |
| Author |
Message |
Shining Arcanine
Veteran
Joined: 24 Sep 2009
Posts: 1110
|
|
| Back to top |
|
 |
albright
Advocate
Joined: 16 Nov 2003
Posts: 2588
Location: Near Toronto
|
 Posted: Sat Feb 27, 2010 2:37 am Post subject: |
 |
|
the zen kernel uses at least some of his patches (BFS for sure);
there's a thread nearby about it
_________________
.... there is nothing - absolutely nothing - half so much worth
doing as simply messing about with Linux ...
(apologies to Kenneth Graeme) |
|
| Back to top |
|
|
mikegpitt
Advocate
Joined: 22 May 2004
Posts: 3224
|
| Posted: Sat Feb 27, 2010 2:42 am Post subject: |
|
|
The ck-sources used to be in the tree years ago, but I guess the patches went away for a while and left the tree.
There seems to be a bug open here to add the ebuild again:
https://bugs.gentoo.org/show_bug.cgi?id=297169 |
|
| Back to top |
|
|
loftwyr
l33t
Joined: 29 Dec 2004
Posts: 970
Location: 43°38'23.62"N 79°27'8.60"W
|
| Posted: Sat Feb 27, 2010 3:17 pm Post subject: |
|
|
I'm using them now and have been using the BFS for a while now.
Don't set the frequency too high as it makes everything unstable but the rest of the patches have (it seems to me) a benefit in responsiveness.
_________________
My emerge --info
Have you run revdep-rebuild lately? It's in gentoolkit and it's worth a shot if things don't work well.
Celebrating 5 years of Gentoo-ing. |
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6090
Location: Removed by Neddy
|
| Posted: Sat Feb 27, 2010 7:02 pm Post subject: |
|
|
| mikegpitt wrote: | The ck-sources used to be in the tree years ago, but I guess the patches went away for a while and left the tree.
There seems to be a bug open here to add the ebuild again:
https://bugs.gentoo.org/show_bug.cgi?id=297169 |
I have updated the ebuilds in that bug for 2.6.32 and 2.6.33 and both work.
time to check if the patches work  ( I have hte ck options, 10kHz tick is tempting just for lulz)
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0;
Last edited by Naib on Sat Feb 27, 2010 8:03 pm; edited 1 time in total |
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6090
Location: Removed by Neddy
|
| Posted: Sat Feb 27, 2010 8:03 pm Post subject: |
|
|
ok this is nice, very nice
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0; |
|
| Back to top |
|
|
gqman69
n00b
Joined: 11 Jul 2007
Posts: 50
|
| Posted: Sun Feb 28, 2010 5:19 pm Post subject: BFS and CONFIG_HZ |
|
|
I have a Core i7 920, so there's Hyperthreading (4 cores, 2 threads/core).
If I use BFS and CONFIG_HZ=1000 there seem to be a huge responsiveness problem when
the system is under load. everything become jerky and the IO gets really slow.
If I use 300 or even less then everything runs very smooth. This doesn't happen with the usual scheduler.
However, I do not recommend using 1000 for multicore processors. There is some hiccups doing so, especially under load.
But nothing like BFS. |
|
| Back to top |
|
|
Shining Arcanine
Veteran
Joined: 24 Sep 2009
Posts: 1110
|
| Posted: Sun Feb 28, 2010 6:32 pm Post subject: Re: BFS and CONFIG_HZ |
|
|
| gqman69 wrote: | I have a Core i7 920, so there's Hyperthreading (4 cores, 2 threads/core).
If I use BFS and CONFIG_HZ=1000 there seem to be a huge responsiveness problem when
the system is under load. everything become jerky and the IO gets really slow.
If I use 300 or even less then everything runs very smooth. This doesn't happen with the usual scheduler.
However, I do not recommend using 1000 for multicore processors. There is some hiccups doing so, especially under load.
But nothing like BFS. |
At 8 virtual cores, you are starting to approach the area where the default scheduler is better. I have CONFIG_HZ=1000 with vanilla-sources 2.6.33 on a Core Duo T2400 and I have no hiccups. |
|
| Back to top |
|
|
bollucks
l33t
Joined: 27 Oct 2004
Posts: 606
|
| Posted: Sun Feb 28, 2010 10:43 pm Post subject: |
|
|
Actually the -ck patch was already out with 2.6.32.
http://marc.info/?l=linux-kernel&m=126059627831860&w=2
Somehow people seemed to miss it entirely? Anyway all of the -ck patches have been in -zen since they've been available, not just BFS. |
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6090
Location: Removed by Neddy
|
| Posted: Sun Feb 28, 2010 11:01 pm Post subject: |
|
|
some of us don't want to use zensources
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0; |
|
| Back to top |
|
|
kernelOfTruth
Watchman
Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)
|
| Posted: Sun Feb 28, 2010 11:07 pm Post subject: Re: BFS and CONFIG_HZ |
|
|
| gqman69 wrote: | I have a Core i7 920, so there's Hyperthreading (4 cores, 2 threads/core).
If I use BFS and CONFIG_HZ=1000 there seem to be a huge responsiveness problem when
the system is under load. everything become jerky and the IO gets really slow.
If I use 300 or even less then everything runs very smooth. This doesn't happen with the usual scheduler.
However, I do not recommend using 1000 for multicore processors. There is some hiccups doing so, especially under load.
But nothing like BFS. |
try tweaking the rr_interval whether that helps:
| Code: | | echo "3" > /proc/sys/kernel/rr_interval |
or even
| Code: | | echo "1" > /proc/sys/kernel/rr_interval |
_________________
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 |
|
| Back to top |
|
|
Shining Arcanine
Veteran
Joined: 24 Sep 2009
Posts: 1110
|
| Posted: Sun Feb 28, 2010 11:09 pm Post subject: |
|
|
| Naib wrote: |
some of us don't want to use zensources |
I tend to stay away from zen-sources. It introduces far too many changes about which I know nothing. I recently switched to vanilla-sources-2.6.33 for a similar reason, because gentoo-sources' changelog had patches that did not apply to my system. Having a separate ck-sources would be great. |
|
| Back to top |
|
|
cheater1034
Veteran
Joined: 09 Sep 2004
Posts: 1558
|
| Posted: Sun Feb 28, 2010 11:44 pm Post subject: Re: BFS and CONFIG_HZ |
|
|
| kernelOfTruth wrote: | | gqman69 wrote: | I have a Core i7 920, so there's Hyperthreading (4 cores, 2 threads/core).
If I use BFS and CONFIG_HZ=1000 there seem to be a huge responsiveness problem when
the system is under load. everything become jerky and the IO gets really slow.
If I use 300 or even less then everything runs very smooth. This doesn't happen with the usual scheduler.
However, I do not recommend using 1000 for multicore processors. There is some hiccups doing so, especially under load.
But nothing like BFS. |
try tweaking the rr_interval whether that helps:
| Code: | | echo "3" > /proc/sys/kernel/rr_interval |
or even
| Code: | | echo "1" > /proc/sys/kernel/rr_interval |
|
I wouldn't think rr_interval has anything to do with it, and i still dont know about the setting of it to 1, but to each his own 
In the zen desktop profile rr_interval is 3, among other things changed - sched_iso_cpu is also changed to 25 for the desktop profile, particularly beneficial if you're using sched_iso for X
| Shining Arcanine wrote: | | Naib wrote: |
some of us don't want to use zensources |
I tend to stay away from zen-sources. It introduces far too many changes about which I know nothing. I recently switched to vanilla-sources-2.6.33 for a similar reason, because gentoo-sources' changelog had patches that did not apply to my system. Having a separate ck-sources would be great. |
It doesnt introduce many foreign changes, not anymore atleast.
Of course not everybody uses everything, but the ones that use what's added (ex: amd-k10, lirc, aufs, reiser4, dazuko, psx2usb, slqb, ureadahead, smapi, tuxonice, classic rcu, etc.) are really pleased - and the ones that don't aren't affected anyway.
Some stuff is internal too, like zen-notes, daconfig, etc.
Then there's 2.6-ck with cfs fix, and then a bunch of tweaks are added now, and things like autoiso and cfs boost.
But it's not for everybody so I don't mind , but added and not used stuff doesnt cause problems, so i'm not sure why people merge only the branches they use manually - because it's pointless  - I also dont understand staying away from it because it merges stuff you don't use - look at all the stuff you don't use in the upstream kernel
_________________
IRC!: #zen-sources on irc.rizon.net
zen-kernel.org
--
Lost in android development land. |
|
| Back to top |
|
|
bollucks
l33t
Joined: 27 Oct 2004
Posts: 606
|
| Posted: Mon Mar 01, 2010 9:39 am Post subject: |
|
|
| Naib wrote: |
some of us don't want to use zensources |
Oh I wasn't suggesting you should use zen, just pointing out that the patches have been around for a while. |
|
| Back to top |
|
|
j-kidd
Apprentice
Joined: 20 Feb 2003
Posts: 213
|
| Posted: Tue Mar 02, 2010 12:19 am Post subject: |
|
|
| Code: |
- 250 Hz is a good compromise choice allowing server performance
- while also showing good interactive responsiveness even
- on SMP and NUMA systems. If you are going to be using NTSC video
- or multimedia, selected 300Hz instead.
+ 250 HZ is a lousy compromise choice allowing server interactivity
+ while also showing desktop throughput and no extra power saving on
+ laptops. No good for anything.
+
+ Recommend 100 or 1000 instead.
|
 server interactivity + desktop throughput |
|
| Back to top |
|
|
Shining Arcanine
Veteran
Joined: 24 Sep 2009
Posts: 1110
|
| Posted: Tue Mar 02, 2010 3:15 am Post subject: |
|
|
| j-kidd wrote: | | Code: |
- 250 Hz is a good compromise choice allowing server performance
- while also showing good interactive responsiveness even
- on SMP and NUMA systems. If you are going to be using NTSC video
- or multimedia, selected 300Hz instead.
+ 250 HZ is a lousy compromise choice allowing server interactivity
+ while also showing desktop throughput and no extra power saving on
+ laptops. No good for anything.
+
+ Recommend 100 or 1000 instead.
|
server interactivity + desktop throughput |
When it comes to software design, compromises usually mean that you get to experience the worst of both worlds. |
|
| Back to top |
|
|
gqman69
n00b
Joined: 11 Jul 2007
Posts: 50
|
| Posted: Tue Mar 02, 2010 10:01 pm Post subject: Re: BFS and CONFIG_HZ |
|
|
| Shining Arcanine wrote: | | gqman69 wrote: | I have a Core i7 920, so there's Hyperthreading (4 cores, 2 threads/core).
If I use BFS and CONFIG_HZ=1000 there seem to be a huge responsiveness problem when
the system is under load. everything become jerky and the IO gets really slow.
If I use 300 or even less then everything runs very smooth. This doesn't happen with the usual scheduler.
However, I do not recommend using 1000 for multicore processors. There is some hiccups doing so, especially under load.
But nothing like BFS. |
At 8 virtual cores, you are starting to approach the area where the default scheduler is better. I have CONFIG_HZ=1000 with vanilla-sources 2.6.33 on a Core Duo T2400 and I have no hiccups. |
Oh I understand you have no "hiccups" because it all depends on what you do. If you try using KDE4 + Composite and compile something that takes 75-100% CPU on all cores you WILL have hiccups and NO it's not "Normal". Multitasking CAN be smooth if you are setup correctly. I have fixed that with CONFIG_HZ=300, now everything smooth all the time including when I my cores are at 100% with "stress -c 8 -i 2" and lets say openoffice is compiled in the background.
My disk access is good, the desktop response if fluid with no jerkiness.
I use Nvidia, EXT4, CFS, with irqbalance and __GL_YIELD=NOTHING. |
|
| Back to top |
|
|
|
Unsupported Software
