From: William Lee Irwin III <wli@holomorphy.com>
To: Mike Galbraith <efault@gmx.de>
Cc: Maciej Soltysiak <solt@dns.toxicfilms.tv>,
"Martin J. Bligh" <mbligh@aracnet.com>,
Andrew Morton <akpm@digeo.com>,
linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: 2.5.70-mm6
Date: Tue, 10 Jun 2003 12:11:47 -0700 [thread overview]
Message-ID: <20030610191147.GC26348@holomorphy.com> (raw)
In-Reply-To: <5.2.0.9.2.20030610193426.00cd9528@pop.gmx.net>
At 04:41 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>> \vomit{Next you'll be telling me worse is better.}
On Tue, Jun 10, 2003 at 08:53:51PM +0200, Mike Galbraith wrote:
> That's an unearned criticism.
> Timeslice management is currently an approximation. IFF the approximation
> is good enough, it will/must out perform pedantic bean-counting. Current
> timeslice management apparently isn't quite good enough, so I'm trying to
> find a way to increase it's informational content without the (in general
> case useless) overhead of attempted perfection. I'm failing miserably btw
> ;-)
The criticism was of the maxim invoked, not the specific direction
you're hacking in.
At 04:41 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>> The issue is the driver returning garbage; not having as good of
>> precision from hardware is no fault of the method. I'd say timer_pit
>> should just return jiffies converted to nanoseconds.
On Tue, Jun 10, 2003 at 08:53:51PM +0200, Mike Galbraith wrote:
> That's the option that I figured was April 1 material, because of the
> missing precision. If it could be made (somehow that I don't understand)
> to produce a reasonable approximation of a high resolution clock, sure.
If there's a TSC, it can be used for extra interpolation. But I think
timer_tsc already does that. I don't think it's quite so laughable; the
machines missing reliable time sources are truly crippled in some way,
by obsolescence or misdesign. I wouldn't call this a deficit. The
major platforms will do fine, and the rest will do no worse than now
apart from perhaps some function call and arithmetic overhead.
At 04:41 AM 6/10/2003 -0700, William Lee Irwin III wrote:
>> Also, I posted the "thud" fix earlier in this thread in addition to the
>> monotonic_clock() bits. AFAICT it mitigates (or perhaps even fixes) an
>> infinite priority escalation scenario.
On Tue, Jun 10, 2003 at 08:53:51PM +0200, Mike Galbraith wrote:
> (yes, mitigates... maybe cures with round down, not really sure)
> Couple that change with reintroduction of backboost to offset some of it's
> other effects and a serious reduction of CHILD_PENALTY and you'll have a
> very snappy desktop. However, there is another side of the
> equation. Large instantaneous variance in sleep_avg/prio also enables the
> scheduler to react quickly such that tasks which were delayed for whatever
> reason rapidly get a chance collect their ticks and catch up. It can and
> does cause perceived unfairness... which is in reality quite fair. It's
> horrible for short period concurrency, but great for long period
> throughput. AFAIKT, it's a hard problem.
I don't know that there are answers better than mitigation.
I propose we err on the other side of the fence and back off as cases
where the larger instantaneous variances are more clearly needed arise.
-- wli
next prev parent reply other threads:[~2003-06-10 19:00 UTC|newest]
Thread overview: 43+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-06-07 22:14 2.5.70-mm6 Andrew Morton
2003-06-08 0:37 ` 2.5.70-mm6 Alexander Hoogerhuis
2003-06-08 0:56 ` 2.5.70-mm6 Andrew Morton
2003-06-08 1:13 ` 2.5.70-mm6 Alexander Hoogerhuis
2003-06-08 12:25 ` 2.5.70-mm6 Christoph Hellwig
2003-06-08 12:09 ` 2.5.70-mm6 Felipe Alfaro Solana
2003-06-08 15:56 ` [patch] 2.5.70-mm6: ethertap.c doesn't compile Adrian Bunk
2003-06-08 16:00 ` Arnaldo Carvalho de Melo
2003-06-08 21:41 ` Adrian Bunk
2003-06-08 22:52 ` 2.5.70-mm6 Pasi Savolainen
2003-06-08 23:17 ` [patch] 2.5.70-mm6: isp driver cleanups Adrian Bunk
2003-06-08 23:27 ` Andrew Morton
2003-06-08 23:24 ` 2.5.70-mm6 Joe
2003-06-09 17:45 ` 2.5.70-mm6 Maciej Soltysiak
2003-06-09 17:39 ` 2.5.70-mm6 Martin J. Bligh
2003-06-09 18:19 ` 2.5.70-mm6 Maciej Soltysiak
2003-06-09 18:51 ` 2.5.70-mm6 Martin J. Bligh
2003-06-09 19:42 ` 2.5.70-mm6 Maciej Soltysiak
2003-06-09 20:04 ` 2.5.70-mm6 William Lee Irwin III
2003-06-10 8:54 ` 2.5.70-mm6 Maciej Soltysiak
2003-06-10 9:20 ` 2.5.70-mm6 William Lee Irwin III
2003-06-10 11:31 ` 2.5.70-mm6 Mike Galbraith
2003-06-10 11:41 ` 2.5.70-mm6 William Lee Irwin III
2003-06-10 18:53 ` 2.5.70-mm6 Mike Galbraith
2003-06-10 19:11 ` William Lee Irwin III [this message]
2003-06-10 11:52 ` 2.5.70-mm6 Maciej Soltysiak
2003-06-09 20:08 ` 2.5.70-mm6 Felipe Alfaro Solana
2003-06-09 20:14 ` 2.5.70-mm6 Martin J. Bligh
2003-06-09 21:09 ` 2.5.70-mm6 Felipe Alfaro Solana
2003-06-10 9:56 ` 2.5.70-mm6 Felipe Alfaro Solana
2003-06-09 18:39 ` 2.5.70-mm6 Felipe Alfaro Solana
2003-06-09 18:06 ` 2.5.70-mm6 Alistair J Strachan
2003-06-09 18:36 ` 2.5.70-mm6 Felipe Alfaro Solana
2003-06-09 19:07 ` 2.5.70-mm6 Pasi Savolainen
2003-06-09 21:20 ` 2.5.70-mm6 Diego Calleja García
2003-06-09 21:40 ` 2.5.70-mm6 Andrew Morton
2003-06-10 12:10 ` 2.5.70-mm6 Grzegorz Jaskiewicz
2003-06-11 2:15 ` 2.5.70-mm6 Mingming Cao
2003-06-11 3:12 ` 2.5.70-mm6 Andrew Morton
2003-06-11 22:15 ` 2.5.70-mm6 Mingming Cao
2003-06-12 16:37 ` 2.5.70-mm6 Mingming Cao
2003-06-12 17:50 ` 2.5.70-mm6 Andrew Morton
2003-06-12 18:43 ` 2.5.70-mm6 Mingming Cao
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=20030610191147.GC26348@holomorphy.com \
--to=wli@holomorphy.com \
--cc=akpm@digeo.com \
--cc=efault@gmx.de \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mbligh@aracnet.com \
--cc=solt@dns.toxicfilms.tv \
/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).