| View previous topic :: View next topic |
| Author |
Message |
fallow
Bodhisattva
Joined: 08 Jan 2004
Posts: 2208
Location: Poland
|
 Posted: Mon Aug 09, 2004 9:31 am Post subject: |
 |
|
Yes , I wrote " little less " - promotions in mind .
| Peter Williams about spa_sc wrote: |
Slightly modified version of Con Kolivas's staircase O(1) scheduler
with active/expired arrays replaced by a single array and an O(1)
promotion mechanism. |
also new zapchod is good
| Peter Willimas about zapchod wrote: |
The following table of hackbench results for various configurations (IAB column is the maximum allowed interactive bonus and TPB is the maximum allowed throughput bonus) illustrates this scheduler's performance compared to the normal O(1) scheduler.
Kernel Mode IAB TPB 10 groups 100 groups
------------------------------------------------------------
vanilla 1.583(100.0%, 2.2%) 16.740(100.0%, 1.5%)
zaphod "eb" 10 5 1.612(101.8%, 1.0%) 16.422( 98.1%, 1.4%)
zaphod "pb" 10 5 1.570( 99.1%, 0.8%) 16.100( 96.1%, 0.7%)
zaphod "eb" 0 0 1.575( 99.5%, 0.1%) 15.992( 95.5%, 1.0%)
zaphod "pb" 0 0 1.602(101.2%, 5.0%) 15.983( 95.4%, 0.1%)
zaphod "eb" 10 0 1.612(101.8%, 0.1%) 16.142( 96.4%, 0.3%)
zaphod "pb" 10 0 1.584(100.0%, 0.1%) 16.014( 95.6%, 0.01%)
zaphod "eb" 0 5 1.609(101.6%, 0.4%) 16.487( 98.4%, 0.9%)
zaphod "pb" 0 5 1.606(101.4%, 0.5%) 16.367( 97.7%, 1.3%)
The numbers are the results of averaging over 10 runs with the numbers in brackets being the size of the value compared to that for the vanilla kernel and the standard deviation of the value as a percentage. The results for 10 groups show there's no significant difference between the schedulers with the differences being mostly less than half a standard deviation. However, for 100 groups several of the results have differences greater than 1 standard deviation.
Other scheduling parameters were left with their default values but further experimentation will be done to determine their efficacy.
|
greetings 
ps.my own small wish is , to do choice beetween vesa rrc and tng , I really don`t like tng and have many of troubles with it.
_________________
"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  |
|
| Back to top |
|
 |
eldiablo
Tux's lil' helper
Joined: 20 Sep 2003
Posts: 136
Location: Halsa(Where keiko lived his last days), Norway
|
| Posted: Mon Aug 09, 2004 10:13 am Post subject: |
|
|
| xiphux wrote: |
I don't think I'm going to be making an actual release soon, especially with all the cleaning and reworking I'm doing. But I plan to commit an almost-complete version of the redone tree to CVS soon.
|
That's as good as a release for me
_________________
Always running the most experimental |
|
| Back to top |
|
|
xiphux
Apprentice
Joined: 15 Jun 2002
Posts: 225
Location: Madison, WI
|
| Posted: Mon Aug 09, 2004 5:54 pm Post subject: |
|
|
That quote about spa-sc doesn't say anything. Spa-sc doesn't have less promotions than regular staircase, it has the same. They are both O(1) schedulers since they operate with scalar values for calculations (all those constants defined at the beginning of sched.c). Spa uses all those promotion interval and max bonus defines, staircase uses the RR_INTERVAL defines, etc. Both spa_sc and staircase call the same calculation functions every scheduler tick. And they are both single array schedulers...
I agree that zaphod is good and worth including, though.
The busted version in cvs has a stock/rrc/tng vesafb option. I'm working on re-adding that into the cleaned up version; the problem is that the actual version of the option was written right into cvs, so I don't have patches on hand. I have to sort through the files in the cvs tree to find the changes I made. |
|
| Back to top |
|
|
xiphux
Apprentice
Joined: 15 Jun 2002
Posts: 225
Location: Madison, WI
|
| Posted: Wed Aug 11, 2004 6:53 am Post subject: |
|
|
Ok, I'm in the middle of committing a cleaned up version of the tree. It has most of the basics... nicksched, staircase, cfqionice, elv-select, etc. I haven't merged the more complicated stuff yet like swsusp2, selectable vesafb rrc/tng, selectable stable/experimental bootsplash, etc. That'll come next. But there are still quite a few differences from the broken tree in there now, so you might want to wait a couple hours before trying to checkout.
It's based on 2.6.8-rc4-mm1. Unfortunately, there are still things that aren't quite right with it: Nicksched and staircase work, but I think I did something to Xsched to make it lock during boot, so I need to investigate that. CFQ with elv-select works, but cfqionice has a couple issues. It works fine for the most part, but sometimes it will oops out and lock the machine... I have to look at the ioprio stuff again. I had enough trouble fixing the reiser4 snapshot and other things to work with the newest -mm... but I know how it gets while waiting for a new version, so I figured I'd commit something that you guys could at least play with temporarily. I'll also commit to cvs more frequently now. This last one just took so long because I started from scratch and wanted to wait until I had something at least partially working.
In a couple hours, I'm also going to add another module to cvs. I'll try to keep an up-to-date collection of the patches that are currently used in there. So you'll be able to check out module xx-sources for the tree, and module xx-patches for the broken-out collection.
[edit]
The cfqionice issue is fixed. Added gensplash for now.
[/edit]
[edit2]
Xsched is fixed. Spa-zaphod is in, though I think it may be broken for some people.
[/edit2] |
|
| Back to top |
|
|
yngwin
Retired Dev
Joined: 19 Dec 2002
Posts: 4572
Location: Suzhou, China
|
| Posted: Sat Aug 14, 2004 1:29 pm Post subject: |
|
|
Now that 2.6.8 is released it would be nice to have a new xx-sources release I think, even if not all the options work... We have to keep up our bleeding-edge image, don't we?
_________________
"Those who deny freedom to others deserve it not for themselves." - Abraham Lincoln
Free Culture | Defective by Design | EFF |
|
| Back to top |
|
|
eldiablo
Tux's lil' helper
Joined: 20 Sep 2003
Posts: 136
Location: Halsa(Where keiko lived his last days), Norway
|
| Posted: Sat Aug 14, 2004 6:41 pm Post subject: |
|
|
Even 2.6.8.1 Is out 
_________________
Always running the most experimental |
|
| Back to top |
|
|
Evil Dark Archon
Guru
Joined: 21 Dec 2002
Posts: 562
Location: Santa Rosa, CA
|
| Posted: Sat Aug 14, 2004 9:10 pm Post subject: |
|
|
unfortunately pserver cvs on sourceforge is down, do you think you could make a patch with all the latest changes and put it in the sourceforge files area?
_________________
This post has been over explained for newb-informing purposes.
Registered Linux user 347334
Abit AV8-3rd eye, AMD Athlon64 3500+ 90nm, ATI Radeon x850 pro |
|
| Back to top |
|
|
xiphux
Apprentice
Joined: 15 Jun 2002
Posts: 225
Location: Madison, WI
|
| Posted: Sat Aug 14, 2004 9:58 pm Post subject: |
|
|
Sure, I'll post a patch within the next hour or so. As soon as I finish cleaning/diffing.
Right now, I have it based on 2.6.8.1, plus almost all of the stuff in 2.6.8-rc4-mm1 (since Andrew hasn't been around to push patches upstream).
Unfortunately, there are still a couple things missing that were in the old one. Unless anyone has any real desire for something immediately, I'll work on reimplementing the selectable vesafb types next.
I'll put it into cvs too, but there may be a couple bugs... since cvs is down, it doesn't get the time to iron itself out...
[edit]
Patch is up, cvs will be committed later tonight.
[/edit] |
|
| Back to top |
|
|
Evil Dark Archon
Guru
Joined: 21 Dec 2002
Posts: 562
Location: Santa Rosa, CA
|
| Posted: Sun Aug 15, 2004 9:59 am Post subject: |
|
|
got this compile error
| Code: | net/ipv4/route.c: In function `ip_route_input_mc':
net/ipv4/route.c:1477: error: too many arguments to function `fib_validate_source'
net/ipv4/route.c: In function `ip_route_input_slow':
net/ipv4/route.c:1655: error: too many arguments to function `fib_validate_source'
net/ipv4/route.c:1683: error: too many arguments to function `fib_validate_source'
net/ipv4/route.c:1766: error: too many arguments to function `fib_validate_source'
make[2]: *** [net/ipv4/route.o] Error 1
make[1]: *** [net/ipv4] Error 2
make: *** [net] Error 2
|
_________________
This post has been over explained for newb-informing purposes.
Registered Linux user 347334
Abit AV8-3rd eye, AMD Athlon64 3500+ 90nm, ATI Radeon x850 pro |
|
| Back to top |
|
|
xiphux
Apprentice
Joined: 15 Jun 2002
Posts: 225
Location: Madison, WI
|
| Posted: Sun Aug 15, 2004 10:08 am Post subject: |
|
|
| With the patch or cvs? |
|
| Back to top |
|
|
izomorfix
Tux's lil' helper
Joined: 18 Jun 2003
Posts: 81
Location: Romania
|
| Posted: Sun Aug 15, 2004 2:21 pm Post subject: |
|
|
Have you been able to adapt the reiser4-swsusp2-patch to work with swsusp .103 and the latest snapshot of reiser4. I'm not very experienced in kernel development, so it gives me some errors like "DECLARE_SWSUSP_LOCAL VARS not defined ....." etc. something must have changed but i don't know where. Please help
_________________
0100100100100000011011000110111101110110
0110010100100000010001110110010101101110
01110100011011110110111100100001 |
|
| Back to top |
|
|
xiphux
Apprentice
Joined: 15 Jun 2002
Posts: 225
Location: Madison, WI
|
| Posted: Sun Aug 15, 2004 6:24 pm Post subject: |
|
|
I dug up the old reiser4-swsusp patch that I modified before, and it applies cleanly after adding swsusp2 .103 and fixing the rejects. (That'll be making its way into cvs, by the way). It compiles without error for me, but I don't use suspend so I don't know if it works correctly - I don't see why it shouldn't, though. I can post the modified patch, if you want it.
[edit]
err... scratch that, there are some errors. I can fix them, though.
[/edit]
[edit2]
Are you sure you even need the reiser4 swsusp patch anymore? A few things from swsusp2 are trickling into the big swsusp1 overhaul they're doing now, so if reiser4 is keeping up with the mm's, it may already be swsusp2-compatible.
[/edit2] |
|
| Back to top |
|
|
Evil Dark Archon
Guru
Joined: 21 Dec 2002
Posts: 562
Location: Santa Rosa, CA
|
| Posted: Sun Aug 15, 2004 9:22 pm Post subject: |
|
|
that compile error was with the patch, pserver cvs is still down
_________________
This post has been over explained for newb-informing purposes.
Registered Linux user 347334
Abit AV8-3rd eye, AMD Athlon64 3500+ 90nm, ATI Radeon x850 pro |
|
| Back to top |
|
|
xiphux
Apprentice
Joined: 15 Jun 2002
Posts: 225
Location: Madison, WI
|
| Posted: Sun Aug 15, 2004 9:39 pm Post subject: |
|
|
| Oh... I compiled that same file and didn't have any trouble... can I see your config? |
|
| Back to top |
|
|
Evil Dark Archon
Guru
Joined: 21 Dec 2002
Posts: 562
Location: Santa Rosa, CA
|
| Posted: Mon Aug 16, 2004 12:51 am Post subject: |
|
|
http://www.andrewnelson.org/.config
_________________
This post has been over explained for newb-informing purposes.
Registered Linux user 347334
Abit AV8-3rd eye, AMD Athlon64 3500+ 90nm, ATI Radeon x850 pro |
|
| Back to top |
|
|
xiphux
Apprentice
Joined: 15 Jun 2002
Posts: 225
Location: Madison, WI
|
| Posted: Mon Aug 16, 2004 1:55 am Post subject: |
|
|
That's weird... I started with a fresh 2.6.8.1 tree, patched it with the rolled up -xx1 patch, loaded up your config, and compiled straight through without trouble. net/ipv4/route.c definitely compiled without even a warning...
What gcc are you using? |
|
| Back to top |
|
|
Evil Dark Archon
Guru
Joined: 21 Dec 2002
Posts: 562
Location: Santa Rosa, CA
|
| Posted: Mon Aug 16, 2004 2:53 am Post subject: |
|
|
3.4.1-r2, i'll try it with 3.3.4-r1.
*UPDATE* i found the problem, obsolete version of ip_fib.h, a leftover from 2.6.8-rc4-evil1 (evil-sources is based on xx-sources).
_________________
This post has been over explained for newb-informing purposes.
Registered Linux user 347334
Abit AV8-3rd eye, AMD Athlon64 3500+ 90nm, ATI Radeon x850 pro |
|
| Back to top |
|
|
yngwin
Retired Dev
Joined: 19 Dec 2002
Posts: 4572
Location: Suzhou, China
|
| Posted: Tue Aug 17, 2004 12:30 pm Post subject: |
|
|
I get this on compiling 2.6.8.1-xx1:
| Code: | LD drivers/net/built-in.o
CC [M] drivers/net/via-rhine.o
drivers/net/via-rhine.c: In function `rhine_init_one':
drivers/net/via-rhine.c:820: error: `CmdReset' undeclared (first use in this function)
drivers/net/via-rhine.c:820: error: (Each undeclared identifier is reported only once
drivers/net/via-rhine.c:820: error: for each function it appears in.)
drivers/net/via-rhine.c:822: warning: implicit declaration of function `wait_for_reset'
drivers/net/via-rhine.c:822: error: `shortname' undeclared (first use in this function)
drivers/net/via-rhine.c:832: warning: implicit declaration of function `reload_eeprom'
make[2]: *** [drivers/net/via-rhine.o] Error 1
make[1]: *** [drivers/net] Error 2
make: *** [drivers] Error 2 |
It's not fatal for me, because I don't normally use the via-rhine card.
_________________
"Those who deny freedom to others deserve it not for themselves." - Abraham Lincoln
Free Culture | Defective by Design | EFF |
|
| Back to top |
|
|
yngwin
Retired Dev
Joined: 19 Dec 2002
Posts: 4572
Location: Suzhou, China
|
| Posted: Tue Aug 17, 2004 12:37 pm Post subject: |
|
|
I also get:
| Code: | CC net/ipv4/sysctl_net_ipv4.o
CC net/ipv4/fib_frontend.o
net/ipv4/fib_frontend.c: In function `fib_validate_source':
include/net/ip_fib.h:182: sorry, unimplemented: inlining failed in call to 'fib_result_table': function body not available
net/ipv4/fib_frontend.c:232: sorry, unimplemented: called from here
include/net/ip_fib.h:182: sorry, unimplemented: inlining failed in call to 'fib_result_table': function body not available
net/ipv4/fib_frontend.c:244: sorry, unimplemented: called from here
make[2]: *** [net/ipv4/fib_frontend.o] Error 1
make[1]: *** [net/ipv4] Error 2
make: *** [net] Error 2 |
Could this be because I use gcc-3.4.1?
_________________
"Those who deny freedom to others deserve it not for themselves." - Abraham Lincoln
Free Culture | Defective by Design | EFF |
|
| Back to top |
|
|
yngwin
Retired Dev
Joined: 19 Dec 2002
Posts: 4572
Location: Suzhou, China
|
| Posted: Tue Aug 17, 2004 12:59 pm Post subject: |
|
|
So it seems. Using gcc-3.3.2 it compiled fine. Now see what happens when I reboot.
Any thoughts on the new Reiser4 FS format, concerning the warning at the end of ftp://ftp.namesys.com/pub/reiser4progs/README ?
_________________
"Those who deny freedom to others deserve it not for themselves." - Abraham Lincoln
Free Culture | Defective by Design | EFF |
|
| Back to top |
|
|
izomorfix
Tux's lil' helper
Joined: 18 Jun 2003
Posts: 81
Location: Romania
|
| Posted: Tue Aug 17, 2004 3:14 pm Post subject: |
|
|
I'm not able to see that message but, if it's the superblock backup change, don't worry about it. the debug + fsck doesn't harm your fs.
to xx:Anyway, I tried to patch the mm-kernel with reiser and swsusp, fixed all the errors ,compiled ,rebooted, and just when I tried to hibernate, BAM, a power surge. Lucky me  , after that when I rebooted I got zam-[1476]: tx block at 398376 corrupted.
KERNEL PANIC  Oh , I was so happy, I thought goodbye all the data hello outside world , After this I fixed my fs using knoppix and the reiser4progs I compiled on it, and voila, just 200 files zeroed, nothing to worry about. The next day I encouraged myself to press enter after i've typed hibernate. Guess what: On restart, no blocks corrupted (so it was the power surge hmmmm. they should've fixed that by now), but the damn udevd (028) doesn't handle suspending or something coz a lot of error messages appeared. udevd ...... can't handle paging request.
OK , now I'm going to my computer to remake my files. I'll be away(probably)
_________________
0100100100100000011011000110111101110110
0110010100100000010001110110010101101110
01110100011011110110111100100001 |
|
| Back to top |
|
|
yngwin
Retired Dev
Joined: 19 Dec 2002
Posts: 4572
Location: Suzhou, China
|
|
| Back to top |
|
|
yngwin
Retired Dev
Joined: 19 Dec 2002
Posts: 4572
Location: Suzhou, China
|
| Posted: Tue Aug 17, 2004 7:44 pm Post subject: |
|
|
The ebuild for xx-sources-2.6.8.1-r1 has been uploaded and is available on the sourceforge page.
_________________
"Those who deny freedom to others deserve it not for themselves." - Abraham Lincoln
Free Culture | Defective by Design | EFF |
|
| Back to top |
|
|
Gentree
Watchman
Joined: 01 Jul 2003
Posts: 5350
Location: France, Old Europe
|
| Posted: Tue Aug 17, 2004 9:28 pm Post subject: |
|
|
Yeah . Incompatiblitly is a pain but it's still only test release so we cant moan.
I have a perfectly reliable reiser4 part. done around march time, I can no longer fsck it with the new tools , nor can I create an new r4 to use with my 2.6.6-love4 kerenl since the reiser4 support in there is the old format and and thinks the new partitions are bad.
I dont think there is much more to it than that, but that what its about from a user's point of view.
That is basically why I am looking at this thread . Newer love stuff seems a bit iffy on r4 and I am looking to see if xx will make a good replacement kernel.
My main requirements are : reiser4 , bootsplash, nvidia gfx, vesa frame buffer modes. If I also have NickShed and cfq it will save me some messing .
Anything I should know before attempting a switch to xx ?
Thx
_________________
Linux, because I'd rather own a free OS than steal one that's not worth paying for.
Gentoo because I'm a masochist
AthlonXP-M on A7N8X. Portage ~x86 |
|
| Back to top |
|
|
Evil Dark Archon
Guru
Joined: 21 Dec 2002
Posts: 562
Location: Santa Rosa, CA
|
| Posted: Tue Aug 17, 2004 10:13 pm Post subject: |
|
|
xx has gensplash, which if i'm correct is compatible with bootsplash themes, there is a howto on the forums somewhere, and it has all the other things you mentioned, for the cpu scheduler you can select between the default, staircase, nicksched, and some others that i don't remember off the top of my head, and it works great, i don't think that you will have a problem with the switch to -xx.
_________________
This post has been over explained for newb-informing purposes.
Registered Linux user 347334
Abit AV8-3rd eye, AMD Athlon64 3500+ 90nm, ATI Radeon x850 pro |
|
| Back to top |
|
|
|
|
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
|
|
Kernel & Hardware
