Skip to contentSkip to conversation

nicksched v30

DISCUSSION Nick Piggin 7 participants 15 messages Mar 5, 2004 – Mar 30, 2004 1 message not in the archive

lore
Share & tools
lore
#1 lore raw
Quoted 6 lines · Nick Piggin

http://www.kerneltrap.org/~npiggin/v30.gz

Applies to kernel 2.6.4-rc1-mm2.
Run X at about nice -10 or -15.
Please report interactivity problems with the default scheduler before using this one etc etc.

So far i noticed:

with default scheduler:

When I run emerge sync (I am on gentoo) and finally the cache on disk gets updated (very heavy disk activity), default scheduler (in conjunction with cfq, haven' tried other) causes a schmall pause of 1-2 seconds when I use my browser, ie mouse cursor is ok, but I cannot scroll for that time.

your scheduler: I tried it with the "love-sources", so maybe that patch "steel300" incorporated was already a bit outdated, but I dunno. It had the same version as above. When I click on a link in thunderbird and it opens up in firefox (is started and just a new tab is created) the mouses stutters for a brief moment. This didn't happen with default scheduler. I dunno if above sympton happens with yours.

I haven't reniced X withy our patch. But I did it once and I didn't see much of a difference. (I dunno if the mouse stutter went away then...)

bye,

Prakash

#2 lore raw
Quoted 10 lines · Prakash K. Cheemplavam
Quoted 6 lines · Nick Piggin

http://www.kerneltrap.org/~npiggin/v30.gz

Applies to kernel 2.6.4-rc1-mm2.
Run X at about nice -10 or -15.
Please report interactivity problems with the default scheduler before using this one etc etc.

Quoted 6 lines

So far i noticed:

with default scheduler:

When I run emerge sync (I am on gentoo) and finally the cache on disk gets updated (very heavy disk activity), default scheduler (in conjunction with cfq, haven' tried other) causes a schmall pause of 1-2 seconds when I use my browser, ie mouse cursor is ok, but I cannot scroll for that time.

Probably just reading stuff in.

Quoted 4 lines

your scheduler: I tried it with the "love-sources", so maybe that patch "steel300" incorporated was already a bit outdated, but I dunno. It had the same version as above. When I click on a link in thunderbird and it opens up in firefox (is started and just a new tab is created) the mouses stutters for a brief moment. This didn't happen with default scheduler. I dunno if above sympton happens with yours.

I haven't reniced X withy our patch. But I did it once and I didn't see much of a difference. (I dunno if the mouse stutter went away then...)

If you don't renice X then it doesn't get any special treatment and interactivity is generally worse. It makes a very big difference here when I run X at -15.

#4 lore raw
Quoted 4 lines · Markus Hästbacka
Quoted 2 lines · Nick Piggin

http://www.kerneltrap.org/~npiggin/v30.gz

Quoted 8 lines
Quoted 6 lines

Applies to kernel 2.6.4-rc1-mm2.
Run X at about nice -10 or -15.
Please report interactivity problems with the default scheduler before using this one etc etc.

Thanks

Same patch for 2.6.4-rc2? mm breaks up a few things for me :(

What does that mean?

#5 lore raw
In reply to Markus Hästbacka
Quoted 4 lines · Markus Hästbacka
Quoted 2 lines · Nick Piggin

http://www.kerneltrap.org/~npiggin/v30.gz

Quoted 10 lines
Quoted 7 lines

Applies to kernel 2.6.4-rc1-mm2.
Run X at about nice -10 or -15.
Please report interactivity problems with the default scheduler before using this one etc etc.

Thanks

Same patch for 2.6.4-rc2? mm breaks up a few things for me :(

Unfortunately not. The scheduler in -mm is different enough that porting isn't straightfoward.

What does mm break for you?

#6 lore raw
Quoted 4 lines · Nick Piggin

Unfortunately not. The scheduler in -mm is different enough that porting isn't straightfoward.

What does mm break for you?

I don't know about the current, but one or two versions back when I tested the last time it printed something about unknown key for the whole dmesg and couldn't find /dev/hd{b,c}. After investigations the hard-drives had changed place for some reason (Not swapped place, but they were in a weird location, can't recall which).

I'll try the new one when I have time to boot this one.

Markus

#7 lore raw
Quoted 7 lines · Markus Hästbacka
Quoted 5 lines · Nick Piggin

Unfortunately not. The scheduler in -mm is different enough that porting isn't straightfoward.

What does mm break for you?

Quoted 9 lines

I don't know about the current, but one or two versions back when I tested the last time it printed something about unknown key for the whole dmesg and couldn't find /dev/hd{b,c}. After investigations the hard-drives had changed place for some reason (Not swapped place, but they were in a weird location, can't recall which).

I'll try the new one when I have time to boot this one.

That would be good. Most stuff in mm is slated to get into the official tree sooner or later, so finding problems sooner is obviously preferable. Thanks

#8 lore raw
Quoted 5 lines · Nick Piggin

That would be good. Most stuff in mm is slated to get into the official tree sooner or later, so finding problems sooner is obviously preferable. Thanks

I've tried it now, seems to work now, maybe there was some problems in earlier versions, now they seem to be fixed.

And feedback of the scheduler:
The scheduler work as great as always, everything works better, a boost in performance. Thanks again.

Markus

#9 lore raw
In reply to Nick Piggin
Quoted 7 lines · Nick Piggin

http://www.kerneltrap.org/~npiggin/v30.gz

Applies to kernel 2.6.4-rc1-mm2. Run X at about nice -10 or -15. Please report interactivity problems with the default scheduler before using this one etc etc.

Thanks

I noticed you took out the SMT code. Does that mean this version of your scheduler is not useful for SMT machines or that the code is just not needed?

Signature · 2 lines
Glenn Johnson
[email protected]
#10 lore raw
In reply to Nick Piggin

Hi Nick,

could do be so nice and do a rediff against current mm kernel? As this kernel finally seems to solve the higher idle temp issue I was having, I would like to try your scheduler once again.

bye,

Prakash

#11 lore raw
Quoted 1 line · Prakash K. Cheemplavam

could do be so nice and do a rediff against current mm kernel? As this

I have done a rediff of it:
http://www.liquidweb.nl/~arvind/patches/v30.tgz.

The integrity isn't as good as the npiggin's one probably.

Arvind.

Signature · 2 lines
Arvind Autar | GnuPG-Key ID: 336E5788
Key fingerprint = FAB8 B3E5 0059 880A 00B8  C859 350E BBDC 336E 5788
#12 lore raw
Quoted 10 lines · Arvind Autar
Quoted 1 line · Prakash K. Cheemplavam

could do be so nice and do a rediff against current mm kernel? As this

I have done a rediff of it:
http://www.liquidweb.nl/~arvind/patches/v30.tgz.

The integrity isn't as good as the npiggin's one probably.

Sorry, please reach me at this address from now on. Thank you. I'll be doing another nicksched update after sched-domains gets into better shape.

#13 lore raw
Quoted 2 lines · n.v.t n.v.t

I have been having trouble compiling the -mm kernel 2.6.5-rc3-mm1 with v30. Please help.

A friend of mine has been having compile troubles also. I used this patch to make things compile:
http://www.liquidweb.nl/~arvind/patches/compile-fix.gz

* I don't know if this will breaks things, it seems empirical to me.*

Arvind.

Signature · 2 lines
Arvind Autar | GnuPG-Key ID: 336E5788
Key fingerprint = FAB8 B3E5 0059 880A 00B8  C859 350E BBDC 336E 5788

A message that is not in the archive (sent off-list or not yet received): [email protected]

#15 lore raw
Quoted 1 line · Jason Munro

I'm getting 403 forbidden returned from apache at the above URL.

Hi,

I apologize for the inconveniences. The correct url is:
http://www.liquidweb.nl/~arvind/patches/v30.gz

Arvind.

Signature · 2 lines
Arvind Autar | GnuPG-Key ID: 336E5788
Key fingerprint = FAB8 B3E5 0059 880A 00B8  C859 350E BBDC 336E 5788