From: William Lee Irwin III <wli@holomorphy.com>
To: Hugh Dickins <hugh@veritas.com>
Cc: Andrew Morton <akpm@digeo.com>,
dmccr@us.ibm.com, rwhron@earthlink.net,
linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: 2.5.65-mm2 vs 2.5.65-mm3 (full objrmap)
Date: Sun, 23 Mar 2003 00:56:08 -0800 [thread overview]
Message-ID: <20030323085608.GI1350@holomorphy.com> (raw)
In-Reply-To: <Pine.LNX.4.44.0303230723210.1595-100000@localhost.localdomain>
On Sun, Mar 23, 2003 at 08:14:12AM +0000, Hugh Dickins wrote:
> I build with highpte on one machine for testing purposes, I don't have
> enough memory for it actually to be important. I'm almost always
> working with Andrew's trees, so was using shpte when it was in, but
> not since. I like the idea of shpte (and the UKVA idea), I couldn't
> see its constituency - the small processes immediately needed to cow
> all their page tables, and the large ones should have been using
> huge pages instead (or such was my misperception).
There's some recent benchmark data from hrandoz showing shpte is
actually doing very well on the speed front lately.
As far as constituency goes I mostly see unintelligent forking servers
(unfortunately these are all too common and tend not to cooperate by
using hugetlb etc.) and smaller machines wanting various trimmings from
kernel memory consumption benefitting. I personally see 2-3MB of
pagetable memory savings from it with end-user workstation loads (X,
xterms, xmms, web browsers, very little dynamic fork()/exec()'ing etc.),
which IMHO is a substantial reduction of the runtime footprint of the
kernel. It alone also conserves 5-6MB of pte_chains in addition to ptes
w/o objrmap. I've also overheard strong interest from software vendors.
-- wli
WARNING: multiple messages have this Message-ID (diff)
From: William Lee Irwin III <wli@holomorphy.com>
To: Hugh Dickins <hugh@veritas.com>
Cc: Andrew Morton <akpm@digeo.com>,
dmccr@us.ibm.com, rwhron@earthlink.net,
linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: 2.5.65-mm2 vs 2.5.65-mm3 (full objrmap)
Date: Sun, 23 Mar 2003 00:56:08 -0800 [thread overview]
Message-ID: <20030323085608.GI1350@holomorphy.com> (raw)
In-Reply-To: <Pine.LNX.4.44.0303230723210.1595-100000@localhost.localdomain>
On Sun, Mar 23, 2003 at 08:14:12AM +0000, Hugh Dickins wrote:
> I build with highpte on one machine for testing purposes, I don't have
> enough memory for it actually to be important. I'm almost always
> working with Andrew's trees, so was using shpte when it was in, but
> not since. I like the idea of shpte (and the UKVA idea), I couldn't
> see its constituency - the small processes immediately needed to cow
> all their page tables, and the large ones should have been using
> huge pages instead (or such was my misperception).
There's some recent benchmark data from hrandoz showing shpte is
actually doing very well on the speed front lately.
As far as constituency goes I mostly see unintelligent forking servers
(unfortunately these are all too common and tend not to cooperate by
using hugetlb etc.) and smaller machines wanting various trimmings from
kernel memory consumption benefitting. I personally see 2-3MB of
pagetable memory savings from it with end-user workstation loads (X,
xterms, xmms, web browsers, very little dynamic fork()/exec()'ing etc.),
which IMHO is a substantial reduction of the runtime footprint of the
kernel. It alone also conserves 5-6MB of pte_chains in addition to ptes
w/o objrmap. I've also overheard strong interest from software vendors.
-- wli
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org">aart@kvack.org</a>
next prev parent reply other threads:[~2003-03-23 8:45 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-03-21 7:58 2.5.65-mm3 Andrew Morton
2003-03-21 7:58 ` 2.5.65-mm3 Andrew Morton
2003-03-21 10:58 ` 2.5.65-mm3 Alexander Hoogerhuis
2003-03-21 11:05 ` 2.5.65-mm3 Andrew Morton
2003-03-21 11:05 ` 2.5.65-mm3 Andrew Morton
2003-03-21 12:05 ` 2.5.65-mm3 Alexander Hoogerhuis
2003-03-21 12:05 ` 2.5.65-mm3 Alexander Hoogerhuis
2003-03-21 15:23 ` [BUG] 2.5.65-mm3 kernel BUG at fs/ext3/super.c:1795! Alexander Hoogerhuis
2003-03-21 20:39 ` Andrew Morton
2003-03-21 20:39 ` Andrew Morton
2003-03-22 2:55 ` Alexander Hoogerhuis
2003-03-22 2:55 ` Alexander Hoogerhuis
2003-03-21 20:15 ` 2.5.65-mm3 Seth Chandler
2003-03-21 20:15 ` 2.5.65-mm3 Seth Chandler
2003-03-21 20:17 ` 2.5.65-mm3 Robert Love
2003-03-21 20:17 ` 2.5.65-mm3 Robert Love
2003-03-22 12:38 ` 2.5.65-mm3 bad: scheduling while atomic! [SCSI] Alexander Hoogerhuis
2003-03-22 12:38 ` Alexander Hoogerhuis
2003-03-22 14:49 ` Alan Cox
2003-03-22 14:49 ` Alan Cox
2003-03-22 16:06 ` 2.5.65-mm2 vs 2.5.65-mm3 (full objrmap) Martin J. Bligh
2003-03-22 16:06 ` Martin J. Bligh
2003-03-22 23:48 ` Hugh Dickins
2003-03-22 23:48 ` Hugh Dickins
2003-03-23 1:07 ` Martin J. Bligh
2003-03-23 1:07 ` Martin J. Bligh
2003-03-23 8:14 ` Hugh Dickins
2003-03-23 8:14 ` Hugh Dickins
2003-03-23 8:56 ` William Lee Irwin III [this message]
2003-03-23 8:56 ` William Lee Irwin III
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=20030323085608.GI1350@holomorphy.com \
--to=wli@holomorphy.com \
--cc=akpm@digeo.com \
--cc=dmccr@us.ibm.com \
--cc=hugh@veritas.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=rwhron@earthlink.net \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.