From: William Lee Irwin III <wli@holomorphy.com>
To: Linus Torvalds <torvalds@transmeta.com>
Cc: Momchil Velikov <velco@fadata.bg>,
Daniel Phillips <phillips@bonn-fries.net>,
Oliver Xymoron <oxymoron@waste.org>,
Rik van Riel <riel@conectiva.com.br>,
Josh MacDonald <jmacd@CS.Berkeley.EDU>,
linux-kernel <linux-kernel@vger.kernel.org>,
reiserfs-list@namesys.com, reiserfs-dev@namesys.com
Subject: Re: Note describing poor dcache utilization under high memory pressure
Date: Tue, 29 Jan 2002 13:01:16 -0800 [thread overview]
Message-ID: <20020129130116.L899@holomorphy.com> (raw)
In-Reply-To: <20020129123932.K899@holomorphy.com> <Pine.LNX.4.33.0201291240180.1223-100000@penguin.transmeta.com>
In-Reply-To: <Pine.LNX.4.33.0201291240180.1223-100000@penguin.transmeta.com>; from torvalds@transmeta.com on Tue, Jan 29, 2002 at 12:49:24PM -0800
On Tue, Jan 29, 2002 at 12:49:24PM -0800, Linus Torvalds wrote:
> I really isn't a co-incidence. The reason so many architectures have page
> table trees is that most architects try to make good decisions, and a tree
> layout is a simple and efficient data structure that maps well to both
> hardware and to usage patterns.
No debate there.
On Tue, Jan 29, 2002 at 12:49:24PM -0800, Linus Torvalds wrote:
> Hashed page tables are incredibly naive, and perform badly for build-up
> and tear-down (and mostly have horrible cache access patterns). At least
> in some version of the UltraSparc, the Linux tree-based software TLB fill
> outperformed the Solaris version, even though the Solaris version was
> handtuned assembly and used hardware acceleration for the hash
> computations. That should tell you something.
Given this, it appears less useful to play with the representations.
There also appears to be some other material on this subject which
you yourself have written. I'll review that for my own enlightenment,
and regardless, I'll focus on something else.
Thanks again,
Bill
next prev parent reply other threads:[~2002-01-29 21:00 UTC|newest]
Thread overview: 85+ messages / expand[flat|nested] mbox.gz Atom feed top
2002-01-28 17:13 Note describing poor dcache utilization under high memory pressure Josh MacDonald
2002-01-28 17:39 ` Linus Torvalds
2002-01-28 18:01 ` Rik van Riel
2002-01-28 18:21 ` Linus Torvalds
2002-01-28 18:37 ` Rik van Riel
2002-01-28 19:28 ` William Lee Irwin III
2002-01-28 20:01 ` Daniel Phillips
2002-01-28 21:33 ` Rick Stevens
2002-01-28 21:43 ` Rik van Riel
2002-01-28 22:00 ` Rick Stevens
2002-01-28 22:43 ` Daniel Phillips
2002-01-28 23:06 ` Rick Stevens
2002-01-28 23:51 ` [OT] " jepler
2002-01-29 2:30 ` IPmonger
2002-01-29 12:02 ` Karl & Betty Schendel
2002-01-28 22:26 ` Daniel Phillips
2002-01-28 22:34 ` Brian Gerst
2002-01-28 23:08 ` Daniel Phillips
2002-01-28 22:39 ` Daniel Phillips
2002-01-28 23:12 ` Rick Stevens
2002-01-28 23:27 ` Daniel Phillips
2002-01-28 22:01 ` Momchil Velikov
2002-01-28 22:19 ` Daniel Phillips
2002-01-29 1:29 ` Oliver Xymoron
2002-01-29 1:37 ` [reiserfs-list] " Valdis.Kletnieks
2002-01-29 1:45 ` Daniel Phillips
2002-01-29 8:39 ` Momchil Velikov
2002-01-29 8:55 ` Daniel Phillips
2002-01-29 9:20 ` William Lee Irwin III
2002-01-29 9:55 ` Daniel Phillips
2002-01-29 10:18 ` Momchil Velikov
2002-01-29 19:55 ` William Lee Irwin III
2002-01-29 20:08 ` Linus Torvalds
2002-01-29 20:39 ` William Lee Irwin III
2002-01-29 20:49 ` Linus Torvalds
2002-01-29 21:01 ` William Lee Irwin III [this message]
2002-01-29 9:20 ` Momchil Velikov
2002-01-29 10:27 ` Daniel Phillips
2002-01-29 11:54 ` Helge Hafting
2002-01-29 12:33 ` Daniel Phillips
2002-01-30 9:07 ` Horst von Brand
2002-01-30 10:55 ` Daniel Phillips
2002-01-30 14:46 ` Rik van Riel
2002-01-30 14:59 ` Daniel Phillips
2002-01-30 15:54 ` Rik van Riel
2002-01-30 16:34 ` Daniel Phillips
2002-01-29 10:59 ` Rik van Riel
2002-01-29 11:28 ` Daniel Phillips
2002-01-29 11:38 ` Rik van Riel
2002-01-29 12:01 ` Daniel Phillips
2002-01-29 16:57 ` Oliver Xymoron
2002-01-29 17:25 ` Rik van Riel
2002-01-29 20:48 ` Daniel Phillips
2002-01-29 21:00 ` Oliver Xymoron
2002-01-29 21:08 ` Linus Torvalds
2002-01-29 21:13 ` Oliver Xymoron
2002-01-29 21:50 ` Linus Torvalds
2002-01-29 22:02 ` Oliver Xymoron
2002-01-29 22:10 ` Linus Torvalds
2002-01-29 22:53 ` Daniel Phillips
2002-01-29 22:53 ` Daniel Phillips
2002-01-29 23:02 ` Oliver Xymoron
2002-01-29 23:21 ` Daniel Phillips
2002-01-28 19:25 ` [reiserfs-dev] " Hans Reiser
2002-01-28 23:52 ` Daniel Phillips
2002-01-29 0:16 ` Hans Reiser
2002-01-29 0:30 ` Alexander Viro
2002-01-29 10:46 ` Hans Reiser
2002-01-29 14:50 ` Chris Mason
2002-01-29 21:10 ` Hans Reiser
2002-01-30 7:11 ` Oliver Xymoron
2002-01-30 9:57 ` Hans Reiser
2002-01-29 17:28 ` Josh MacDonald
2002-01-29 18:44 ` [reiserfs-list] " Andreas Dilger
2002-01-29 19:55 ` Andrew Morton
2002-01-30 7:17 ` Oliver Xymoron
2002-01-30 7:32 ` [reiserfs-list] Re: [reiserfs-dev] Re: Note describing poordcache " Andrew Morton
2002-01-30 7:52 ` Oliver Xymoron
2002-01-30 10:03 ` Hans Reiser
2002-01-30 10:07 ` [reiserfs-dev] Re: Note describing poor dcache " Horst von Brand
2002-01-29 18:29 ` Horst von Brand
2002-01-29 0:51 ` Daniel Phillips
2002-01-29 1:32 ` Daniel Phillips
2002-01-28 22:46 ` Alex Bligh - linux-kernel
2002-01-29 17:27 ` Josh MacDonald
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=20020129130116.L899@holomorphy.com \
--to=wli@holomorphy.com \
--cc=jmacd@CS.Berkeley.EDU \
--cc=linux-kernel@vger.kernel.org \
--cc=oxymoron@waste.org \
--cc=phillips@bonn-fries.net \
--cc=reiserfs-dev@namesys.com \
--cc=reiserfs-list@namesys.com \
--cc=riel@conectiva.com.br \
--cc=torvalds@transmeta.com \
--cc=velco@fadata.bg \
/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