Skip to contentSkip to conversation

Staircase Deadline cpu scheduler version 0.46

ANNOUNCE Con Kolivas 3 participants 4 messages Apr 22, 2007 – Apr 23, 2007 1 message not in the archive

lore
Share & tools
lore
#0 lore raw

Yet another significant bugfix for SMP balancing was just posted for the staircase deadline cpu scheduler which improves behaviour dramatically on any SMP machine.

Thanks to Willy Tarreau for noticing more bugs.

As requested was a version in the Makefile so this version of the patch adds -sd046 to the kernel version.

http://ck.kolivas.org/patches/staircase-deadline/2.6.21-rc7-sd-0.46.patch
http://ck.kolivas.org/patches/staircase-deadline/2.6.20.7-sd-0.46.patch

Renicing X to -10, while not essential, may be desirable on the desktop. Unlike the CFS scheduler which renices X without your intervention to nice -19, the SD patches do not alter nice level on their own.

See the patch just posted called 'sched: implement staircase deadline scheduler load weight fix' for details of the fixes.

Thanks to all testing and giving feedback.

Well I'm exhausted...

Signature · 1 line
-ck
#1 lore raw

mån 2007-04-23 klockan 01:03 +1000 skrev Con Kolivas:

Quoted 16 lines

Yet another significant bugfix for SMP balancing was just posted for the staircase deadline cpu scheduler which improves behaviour dramatically on any SMP machine.

Thanks to Willy Tarreau for noticing more bugs.

As requested was a version in the Makefile so this version of the patch adds -sd046 to the kernel version.

http://ck.kolivas.org/patches/staircase-deadline/2.6.21-rc7-sd-0.46.patch
http://ck.kolivas.org/patches/staircase-deadline/2.6.20.7-sd-0.46.patch

Renicing X to -10, while not essential, may be desirable on the desktop. Unlike the CFS scheduler which renices X without your intervention to nice -19, the SD patches do not alter nice level on their own.

Quoted 7 lines

See the patch just posted called 'sched: implement staircase deadline scheduler load weight fix' for details of the fixes.

Thanks to all testing and giving feedback.

Well I'm exhausted...

This one broke 2.6.20.7 build...

kernel/sched.c: In function ‘dependent_sleeper’:
kernel/sched.c:3319: error: ‘DEF_TIMESLICE’ undeclared (first use in this function)
kernel/sched.c:3319: error: (Each undeclared identifier is reported only once
kernel/sched.c:3319: error: for each function it appears in.)

Signature · 1 line
Thomas Backlund <[email protected]>
#2 lore raw
Quoted 24 lines · Thomas Backlund

mån 2007-04-23 klockan 01:03 +1000 skrev Con Kolivas:

Quoted 16 lines

Yet another significant bugfix for SMP balancing was just posted for the staircase deadline cpu scheduler which improves behaviour dramatically on any SMP machine.

Thanks to Willy Tarreau for noticing more bugs.

As requested was a version in the Makefile so this version of the patch adds -sd046 to the kernel version.

http://ck.kolivas.org/patches/staircase-deadline/2.6.21-rc7-sd-0.46.patch
http://ck.kolivas.org/patches/staircase-deadline/2.6.20.7-sd-0.46.patch

Renicing X to -10, while not essential, may be desirable on the desktop. Unlike the CFS scheduler which renices X without your intervention to nice -19, the SD patches do not alter nice level on their own.

Quoted 6 lines

See the patch just posted called 'sched: implement staircase deadline scheduler load weight fix' for details of the fixes.

Thanks to all testing and giving feedback.

Well I'm exhausted...

Quoted 8 lines

This one broke 2.6.20.7 build...

kernel/sched.c: In function ‘dependent_sleeper’:
kernel/sched.c:3319: error: ‘DEF_TIMESLICE’ undeclared (first use in this function)
kernel/sched.c:3319: error: (Each undeclared identifier is reported only once
kernel/sched.c:3319: error: for each function it appears in.)

Apologies it was a blind merge.

Use this instead
http://ck.kolivas.org/patches/staircase-deadline/2.6.20.7-sd-0.46-1.patch

Signature · 1 line
-ck
#4 lore raw
In reply to a message not in the archive

From: hechacker1 <[email protected]>
Date: Apr 22, 2007 6:09 PM
Subject: Re: [ck] [ANNOUNCE] Staircase Deadline cpu scheduler version 0.46
To: Con Kolivas <[email protected]>

First of all, thank you for your continued development of SD. I've been using RSDL v.30 since it came out with skunk-sources (a gentoo patchset). It has ran stable since that time.

I went ahead and taught myself how to patch a kernel together so I could try SD-0.46.

hechacker1@700m ~ $ uname -a
Linux 700m 2.6.21-rc7-sd046-rsdl #2 PREEMPT Sun Apr 22 15:13:36 PDT 2007 i686 Intel(R) Pentium(R) M processor 1.70GHz GenuineIntel GNU/Linux

My findings:
SD 0.46 is much more responsive than RSDL 0.30. I have portage niced to 19, make -j2, and it doesn't interfere with Xorg/Beryl at all. Animations are silky smooth in beryl compared to the standard 2.6.21 scheduler which lags/becomes jerky when there is any kind of load (and nicing it didn't help). This is especially important since I use resier4 with cryptcompress (lzo) and it does peg out my CPU at 100% during portage database rebuilds and searches. Not to mention all the untaring/unzipping that portage does to build from source (and i have everything built in tmpfs, so there isn't an i/o limitation).

I have rr_interval at 6 ms.

Xorg is "slightly" more responsive (doesn't lag) under load with nice -10, although I also have beryl and emerald to worry about too since they draw the windows and consume cpu. Xorg with nice 0 is still responsive under all loads, so at least the tweak isn't required.

So far this is my new favorite scheduler. I'm compiling openoffice in the background right now and I can't even notice it.

I will spend some time with this scheduler to get a feel for its performance and later try CFS to get a comparison.

Quoted 30 lines · Con Kolivas

Yet another significant bugfix for SMP balancing was just posted for the
staircase deadline cpu scheduler which improves behaviour dramatically on any SMP machine.

Thanks to Willy Tarreau for noticing more bugs.

As requested was a version in the Makefile so this version of the patch adds -sd046 to the kernel version.

http://ck.kolivas.org/patches/staircase-deadline/2.6.21-rc7-sd-0.46.patch
http://ck.kolivas.org/patches/staircase-deadline/2.6.20.7-sd-0.46.patch

Renicing X to -10, while not essential, may be desirable on the desktop. Unlike the CFS scheduler which renices X without your intervention to nice -19, the SD patches do not alter nice level on their own.

See the patch just posted called 'sched: implement staircase deadline
scheduler load weight fix' for details of the fixes.

Thanks to all testing and giving feedback.

Well I'm exhausted...

--
-ck
_______________________________________________
http://ck.kolivas.org/faqs/replying-to-mailing-list.txt
ck mailing list - mailto: [email protected]
http://vds.kolivas.org/mailman/listinfo/ck