From: Nick Piggin <piggin@cyberone.com.au>
To: bill davidsen <davidsen@tmr.com>
Cc: linux-kernel@vger.kernel.org
Subject: Re: Linux 2.6.0-test6
Date: Thu, 02 Oct 2003 12:45:48 +1000 [thread overview]
Message-ID: <3F7B915C.9050301@cyberone.com.au> (raw)
In-Reply-To: <blfg2i$j43$1@gatekeeper.tmr.com>
bill davidsen wrote:
>In article <3F78D866.5070605@cyberone.com.au>,
>Nick Piggin <piggin@cyberone.com.au> wrote:
>|
>|
>| bill davidsen wrote:
>|
>| >In article <3F77BB2C.7030402@cyberone.com.au>,
>| >Nick Piggin <piggin@cyberone.com.au> wrote:
>| >
>| >| AFAIK, Con's scheduler doesn't change the nice implementation at all.
>| >| Possibly some of his changes amplify its problems, or, more likely they
>| >| remove most other scheduler problems leaving this one noticable.
>| >|
>| >| If X is running at -20, and xmms at +19, xmms is supposed to still get
>| >| 5% of the CPU. Should be enough to run fine. Unfortunately this is
>| >| achieved by giving X very large timeslices, so xmms's scheduling latency
>| >| becomes large. The interactivity bonuses don't help, either.
>| >
>| >Clearly the "some is good, more is better" approach doesn't provide
>| >stable balance between sound and cpu hogs. It isn't a question of "how
>| >much" cpu, just "when"which works or not.
>| >
>| >This is sort of like the deadline scheduler in that it trades of
>| >throughput for avoiding jackpot cases. I think that's desired behaviour
>| >in a CPU schedular too, at least if used by humans.
>| >
>|
>| I'm not sure what you mean. There is nothing good to say about Ingo's
>| nice mechanism though (sorry Ingo, its otherwise a very nice
>| scheduler!).
>
>Oh, I think the test5-mm4 behaviour is far better than yours in terms of
>throughput. I would expect that it could have several percent less
>system time, leaving it for cpu-bound user processes. However, on my
>little test machine (PII-350 w/ 96MB) running patch to get a kernel
>source tree ready makes the system damn near unusable. With your v15
>patch neither patch nor a kernel build is a real problem (I don't use
>-j, there's only one processor and it doesn't help). I can happily read
>mail, run windows to remote machines to do admin, check web based
>monitors, and generally use the system. It's not a ball of fire, but
>it's old and slow moving, and I can identify with that ;-)
>
>| In my scheduler, nice -20 processes get small timeslices so scheduling
>| latency stays low or even gets lower, while nice +19 ones get large
>| timeslices for lower context switches and better cache efficiency. As
>| you would like.
>
>Clearly your timeslices don't get so large the system suffers. And I can
>run setiathome with a little nice, like -5, and it will get some time
>without slowing the important stuff on the system. Even at -19 it does
>keep going. On a small memory machine I hate to give pages to a process
>without giving it the CPU to justify the memory.
>
>I'm going to try test5-n15 against test5-mm4 and test6-std with a server
>load, but for now I run your patch on machines I use as personal
>workstations. I haven't tried it SMP, that day will come.
>
Oh ok, yeah. It does look like it might need some work to prevent
timeslices from getting too small. I'm going to start working on this
soon.
next prev parent reply other threads:[~2003-10-02 2:46 UTC|newest]
Thread overview: 128+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-09-28 1:27 Linux 2.6.0-test6 Linus Torvalds
2003-09-28 7:03 ` Con Kolivas
2003-09-28 10:02 ` Rob Landley
2003-09-29 4:55 ` Nick Piggin
2003-09-29 7:35 ` Rob Landley
2003-09-29 16:55 ` Ed Sweetman
2003-09-30 0:03 ` Nick Piggin
2003-10-02 0:41 ` Pedro Larroy
2003-10-02 3:05 ` Nick Piggin
2003-10-02 19:07 ` Pedro Larroy
2003-10-03 0:07 ` Nick Piggin
2003-10-03 19:34 ` Pedro Larroy
2003-09-29 18:45 ` bill davidsen
2003-09-30 1:12 ` Nick Piggin
2003-10-01 21:13 ` bill davidsen
2003-10-02 2:45 ` Nick Piggin [this message]
2003-09-28 8:26 ` Markus Hästbacka
2003-09-28 10:54 ` Jeff Garzik
2003-09-28 8:59 ` keyboard repeat / sound [was Re: Linux 2.6.0-test6] Roger Luethi
2003-09-29 15:16 ` Vojtech Pavlik
2003-09-30 7:50 ` Paul
2003-09-30 12:51 ` Vojtech Pavlik
2003-09-30 13:21 ` Aristeu Sergio Rozanski Filho
2003-09-30 13:44 ` Vojtech Pavlik
2003-09-30 14:05 ` Aristeu Sergio Rozanski Filho
2003-09-30 14:16 ` Vojtech Pavlik
2003-10-01 23:51 ` Aristeu Sergio Rozanski Filho
2003-09-30 18:16 ` Mark W. Alexander
2003-10-01 23:52 ` Aristeu Sergio Rozanski Filho
2003-09-28 10:09 ` Linux 2.6.0-test6 Rafał 'rmrmg' Roszak
2003-09-28 11:05 ` Andreas Jellinghaus
2003-09-28 12:34 ` Dave Jones
2003-09-28 16:12 ` Andreas Jellinghaus
2003-09-28 17:51 ` Andries Brouwer
2003-09-28 16:42 ` Ivan Gyurdiev
2003-09-28 20:26 ` [patch] 2.6.0-test6: correct hdlcdrv.h prototypes Adrian Bunk
2003-09-29 13:23 ` Linux 2.6.0-test6 Florin Iucha
2003-09-29 13:55 ` Muli Ben-Yehuda
2003-09-29 14:01 ` Jaroslav Kysela
2003-09-29 14:18 ` Muli Ben-Yehuda
2003-09-29 19:04 ` bill davidsen
2003-09-29 14:30 ` Takashi Iwai
2003-09-29 13:58 ` Jaroslav Kysela
2003-09-29 16:30 ` Linux 2.6.0-test6 (compile statistics) John Cherry
2003-09-29 17:44 ` Jesper Juhl
2003-10-06 20:39 ` John Cherry
2003-10-01 8:58 ` Who changed /proc/<pid>/ in 2.6.0-test5-bk9? Mikael Pettersson
2003-10-01 11:52 ` John Levon
2003-10-01 20:21 ` bill davidsen
2003-10-02 1:00 ` John Levon
2003-10-06 3:01 ` bill davidsen
2003-10-01 15:11 ` Linus Torvalds
2003-10-01 20:58 ` bill davidsen
2003-10-01 23:42 ` Albert Cahalan
2003-10-02 0:38 ` Linus Torvalds
2003-10-02 0:57 ` Albert Cahalan
2003-10-02 3:35 ` Ulrich Drepper
2003-10-02 4:12 ` Albert Cahalan
2003-10-02 4:58 ` Ulrich Drepper
2003-10-02 13:48 ` Albert Cahalan
2003-10-02 17:30 ` Ulrich Drepper
2003-10-03 0:03 ` Albert Cahalan
2003-10-03 0:40 ` Linus Torvalds
2003-10-03 2:53 ` Jamie Lokier
2003-10-06 4:54 ` Mike Fedyk
2003-10-06 2:52 ` bill davidsen
2003-10-07 23:08 ` Robert White
2003-10-07 22:46 ` Robert White
2003-10-07 23:25 ` Linus Torvalds
2003-10-08 0:41 ` Robert White
2003-10-08 0:54 ` Linus Torvalds
2003-10-08 2:31 ` Robert White
2003-10-08 2:39 ` David Lang
2003-10-08 2:59 ` Robert White
2003-10-09 18:25 ` bill davidsen
2003-10-08 2:47 ` Who changed /proc/<pid>/ in 2.6.0-test5-bk9? (SIGPIPE?) Robert White
2003-10-08 2:57 ` Linus Torvalds
2003-10-08 4:01 ` Robert White
2003-10-08 4:08 ` Linus Torvalds
2003-10-08 10:47 ` Who changed /proc/<pid>/ in 2.6.0-test5-bk9? bert hubert
2003-10-08 19:12 ` Ulrich Drepper
2003-10-09 18:43 ` bill davidsen
2003-10-08 21:54 ` Robert White
2003-10-09 18:12 ` bill davidsen
2003-10-10 4:39 ` Jamie Lokier
2003-10-09 17:59 ` bill davidsen
2003-10-11 3:02 ` Here is a case that proves my previous position wrong regurading CLONE_THREAD and CLONE_FILES Robert White
2003-10-11 3:48 ` viro
2003-10-12 11:41 ` Jamie Lokier
2003-10-02 8:46 ` Who changed /proc/<pid>/ in 2.6.0-test5-bk9? Miquel van Smoorenburg
2003-10-02 22:35 ` Jamie Lokier
2003-10-02 23:43 ` Miquel van Smoorenburg
2003-10-06 2:57 ` bill davidsen
2003-10-02 3:38 ` Ulrich Drepper
2003-09-28 10:14 Linux 2.6.0-test6 Geert Uytterhoeven
2003-09-28 12:50 ` Russell King
2003-09-28 13:54 ` Bernardo Innocenti
2003-09-28 17:37 ` Linus Torvalds
2003-09-28 18:46 ` Sam Ravnborg
2003-09-28 18:52 ` Linus Torvalds
2003-09-28 19:44 ` Jamie Lokier
2003-09-28 19:16 ` Jörn Engel
2003-09-28 19:31 ` Sam Ravnborg
2003-09-28 19:44 ` Jörn Engel
2003-09-28 19:42 ` Russell King
2003-09-28 20:00 ` Jörn Engel
2003-09-28 21:43 ` Tim Schmielau
2003-09-28 21:54 ` Arnaldo Carvalho de Melo
2003-09-29 15:08 ` Chris Friesen
2003-09-29 15:14 ` Geert Uytterhoeven
2003-09-28 19:28 ` Russell King
2003-09-29 8:52 ` Geert Uytterhoeven
2003-09-29 19:19 ` bill davidsen
2003-09-30 10:42 ` Jörn Engel
2003-09-30 15:17 ` Bill Davidsen
2003-10-01 9:08 ` Jörn Engel
2003-09-28 12:19 Oliver Pitzeier
2003-09-28 12:52 ` Geert Uytterhoeven
2003-09-28 21:41 ` Andrew Morton
2003-09-29 9:58 ` Geert Uytterhoeven
2003-09-29 10:08 ` Russell King
2003-09-29 19:19 ` Anton Blanchard
2003-09-28 12:53 ` viro
2003-09-28 13:54 ` Urban Widmark
2003-09-29 9:00 Oliver Pitzeier
[not found] <<3F7CBDD4.7010503@cyberone.com.au>>
2003-10-03 3:51 ` Jason Munro
2003-10-06 3:05 ` bill davidsen
2003-10-09 18:44 ` bill davidsen
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=3F7B915C.9050301@cyberone.com.au \
--to=piggin@cyberone.com.au \
--cc=davidsen@tmr.com \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).