From: Mike Travis <travis@sgi.com>
To: "H. Peter Anvin" <hpa@zytor.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>,
Christoph Lameter <cl@linux-foundation.org>,
Jeremy Fitzhardinge <jeremy@goop.org>,
Ingo Molnar <mingo@elte.hu>,
Andrew Morton <akpm@linux-foundation.org>,
Jack Steiner <steiner@sgi.com>,
linux-kernel@vger.kernel.org,
Arjan van de Ven <arjan@infradead.org>
Subject: Re: [RFC 00/15] x86_64: Optimize percpu accesses
Date: Thu, 10 Jul 2008 16:37:02 -0700 [thread overview]
Message-ID: <48769D1E.3030003@sgi.com> (raw)
In-Reply-To: <487663C4.6040309@zytor.com>
H. Peter Anvin wrote:
> Mike Travis wrote:
>>
>> The biggest growth came from moving all the xxx[NR_CPUS] arrays into
>> the per cpu area. So you free up a huge amount of unused memory when
>> the NR_CPUS count starts getting into the ozone layer. 4k now, 16k
>> real soon now, ??? future?
>>
>
> Even (or perhaps especially) so, allocating the percpu area in 2 MB
> increments is a total nonstarter. It hurts the small, common
> configurations way too much. For SGI, it's probably fine.
>
> -hpa
Yes, "right-sizing" the kernel for systems from 512M laptops to 4k cpu
systems with memory totally maxed out, using only a binary distribution
has proven tricky (at best... ;-)
One alternative was to only allocate a chunk similar in size to
PERCPU_ENOUGH_ROOM and allow for startup options to create a bigger
space if needed. Though the realloc idea has some merit if we can
validate the non-use of pointers to specific cpu's percpu vars.
(CPU_ALLOC contains a function to dereference a percpu offset.)
The discussion started at:
http://marc.info/?l=linux-kernel&m=121212026716085&w=4
Thanks,
Mike
next prev parent reply other threads:[~2008-07-10 23:37 UTC|newest]
Thread overview: 190+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-07-09 16:51 [RFC 00/15] x86_64: Optimize percpu accesses Mike Travis
2008-07-09 16:51 ` [RFC 01/15] x86_64: Cleanup early setup_percpu references Mike Travis
2008-07-09 16:51 ` [RFC 02/15] x86_64: Fold pda into per cpu area Mike Travis
2008-07-09 22:02 ` Eric W. Biederman
2008-07-13 17:54 ` Ingo Molnar
2008-07-14 14:24 ` Mike Travis
2008-07-09 16:51 ` [RFC 03/15] x86_64: Reference zero-based percpu variables offset from gs Mike Travis
2008-07-09 16:51 ` [RFC 04/15] x86_64: Replace cpu_pda ops with percpu ops Mike Travis
2008-07-09 16:51 ` [RFC 05/15] x86_64: Replace xxx_pda() operations with x86_xxx_percpu() Mike Travis
2008-07-09 16:51 ` [RFC 06/15] x86_64: Replace xxx_pda() operations in include_asm-x86_current_h Mike Travis
2008-07-09 16:51 ` [RFC 07/15] x86_64: Replace xxx_pda() operations in include_asm-x86_hardirq_64_h Mike Travis
2008-07-09 16:51 ` [RFC 08/15] x86_64: Replace xxx_pda() operations in include_asm-x86_mmu_context_64_h Mike Travis
2008-07-09 16:51 ` [RFC 09/15] x86_64: Replace xxx_pda() operations in include_asm-x86_percpu_h Mike Travis
2008-07-09 16:51 ` [RFC 10/15] x86_64: Replace xxx_pda() operations in include_asm-x86_smp_h Mike Travis
2008-07-09 16:51 ` [RFC 11/15] x86_64: Replace xxx_pda() operations in include_asm-x86_stackprotector_h Mike Travis
2008-07-09 16:51 ` [RFC 12/15] x86_64: Replace xxx_pda() operations in include_asm-x86_thread_info_h Mike Travis
2008-07-09 16:51 ` [RFC 13/15] x86_64: Replace xxx_pda() operations in include_asm-x86_topology_h Mike Travis
2008-07-09 16:51 ` [RFC 14/15] x86_64: Remove xxx_pda() operations Mike Travis
2008-07-09 16:51 ` [RFC 15/15] x86_64: Remove cpu_pda() macro Mike Travis
2008-07-09 17:19 ` [RFC 00/15] x86_64: Optimize percpu accesses H. Peter Anvin
2008-07-09 17:40 ` Mike Travis
2008-07-09 17:42 ` H. Peter Anvin
2008-07-09 18:05 ` Mike Travis
2008-07-09 17:44 ` Jeremy Fitzhardinge
2008-07-09 18:09 ` Mike Travis
2008-07-09 18:30 ` H. Peter Anvin
2008-07-09 19:34 ` Ingo Molnar
2008-07-09 19:44 ` H. Peter Anvin
2008-07-09 20:26 ` Adrian Bunk
2008-07-09 21:03 ` Mike Travis
2008-07-09 21:23 ` Jeremy Fitzhardinge
2008-07-25 15:49 ` Mike Travis
2008-07-25 16:08 ` Jeremy Fitzhardinge
2008-07-25 16:46 ` Mike Travis
2008-07-25 16:58 ` Jeremy Fitzhardinge
2008-07-25 18:12 ` Mike Travis
2008-07-09 17:27 ` Jeremy Fitzhardinge
2008-07-09 17:39 ` Christoph Lameter
2008-07-09 17:51 ` Jeremy Fitzhardinge
2008-07-09 18:14 ` Mike Travis
2008-07-09 18:22 ` Jeremy Fitzhardinge
2008-07-09 18:31 ` Mike Travis
2008-07-09 19:08 ` Jeremy Fitzhardinge
2008-07-09 18:02 ` Mike Travis
2008-07-09 18:13 ` Christoph Lameter
2008-07-09 18:26 ` Jeremy Fitzhardinge
2008-07-09 18:34 ` Christoph Lameter
2008-07-09 18:37 ` H. Peter Anvin
2008-07-09 18:48 ` Jeremy Fitzhardinge
2008-07-09 18:53 ` Christoph Lameter
2008-07-09 19:07 ` Jeremy Fitzhardinge
2008-07-09 19:12 ` Christoph Lameter
2008-07-09 19:32 ` Jeremy Fitzhardinge
2008-07-09 19:41 ` Ingo Molnar
2008-07-09 19:45 ` H. Peter Anvin
2008-07-09 19:52 ` Christoph Lameter
2008-07-09 20:00 ` Ingo Molnar
2008-07-09 20:09 ` Jeremy Fitzhardinge
2008-07-09 21:05 ` Mike Travis
2008-07-09 19:44 ` Christoph Lameter
2008-07-09 19:48 ` Jeremy Fitzhardinge
2008-07-09 18:27 ` Mike Travis
2008-07-09 18:46 ` Jeremy Fitzhardinge
2008-07-09 20:22 ` Eric W. Biederman
2008-07-09 20:35 ` Jeremy Fitzhardinge
2008-07-09 20:53 ` Eric W. Biederman
2008-07-09 21:03 ` Ingo Molnar
2008-07-09 21:16 ` H. Peter Anvin
2008-07-09 21:10 ` Arjan van de Ven
2008-07-09 23:20 ` Eric W. Biederman
2008-07-09 18:31 ` H. Peter Anvin
2008-07-09 18:00 ` Mike Travis
2008-07-09 19:05 ` Jeremy Fitzhardinge
2008-07-09 19:28 ` Ingo Molnar
2008-07-09 20:55 ` Mike Travis
2008-07-09 21:12 ` Ingo Molnar
2008-07-09 20:00 ` Eric W. Biederman
2008-07-09 20:05 ` Jeremy Fitzhardinge
2008-07-09 20:15 ` Ingo Molnar
2008-07-09 20:07 ` Ingo Molnar
2008-07-09 20:11 ` Jeremy Fitzhardinge
2008-07-09 20:18 ` Christoph Lameter
2008-07-09 20:33 ` Jeremy Fitzhardinge
2008-07-09 20:42 ` H. Peter Anvin
2008-07-09 20:48 ` Jeremy Fitzhardinge
2008-07-09 21:06 ` Eric W. Biederman
2008-07-09 21:16 ` H. Peter Anvin
2008-07-09 21:20 ` Jeremy Fitzhardinge
2008-07-09 21:25 ` Christoph Lameter
2008-07-09 21:36 ` H. Peter Anvin
2008-07-09 21:41 ` Jeremy Fitzhardinge
2008-07-09 22:22 ` Eric W. Biederman
2008-07-09 22:32 ` Jeremy Fitzhardinge
2008-07-09 23:36 ` Eric W. Biederman
2008-07-10 0:19 ` H. Peter Anvin
2008-07-10 0:24 ` Jeremy Fitzhardinge
2008-07-10 14:14 ` Christoph Lameter
2008-07-10 14:26 ` H. Peter Anvin
2008-07-10 15:26 ` Christoph Lameter
2008-07-10 15:42 ` H. Peter Anvin
2008-07-10 16:24 ` Christoph Lameter
2008-07-10 16:33 ` H. Peter Anvin
2008-07-10 16:45 ` Christoph Lameter
2008-07-10 17:33 ` Jeremy Fitzhardinge
2008-07-10 17:42 ` Christoph Lameter
2008-07-10 17:53 ` Jeremy Fitzhardinge
2008-07-10 17:55 ` H. Peter Anvin
2008-07-10 20:52 ` Christoph Lameter
2008-07-10 20:58 ` Jeremy Fitzhardinge
2008-07-10 21:03 ` H. Peter Anvin
2008-07-11 0:55 ` Mike Travis
2008-07-10 21:05 ` Christoph Lameter
2008-07-10 21:22 ` Eric W. Biederman
2008-07-10 21:29 ` H. Peter Anvin
2008-07-11 0:12 ` Mike Travis
2008-07-11 0:14 ` H. Peter Anvin
2008-07-11 0:58 ` Mike Travis
2008-07-11 1:41 ` H. Peter Anvin
2008-07-11 15:37 ` Christoph Lameter
2008-07-11 0:42 ` Eric W. Biederman
2008-07-11 15:36 ` Christoph Lameter
2008-07-10 17:53 ` H. Peter Anvin
2008-07-10 17:26 ` Eric W. Biederman
2008-07-10 17:38 ` Christoph Lameter
2008-07-10 19:11 ` Mike Travis
2008-07-10 19:12 ` Eric W. Biederman
2008-07-10 17:46 ` Mike Travis
2008-07-10 17:51 ` H. Peter Anvin
2008-07-10 19:09 ` Eric W. Biederman
2008-07-10 19:18 ` Mike Travis
2008-07-10 19:32 ` H. Peter Anvin
2008-07-10 23:37 ` Mike Travis [this message]
2008-07-10 20:17 ` Eric W. Biederman
2008-07-10 20:24 ` Ingo Molnar
2008-07-10 21:33 ` Eric W. Biederman
2008-07-11 1:39 ` Mike Travis
2008-07-11 2:57 ` Eric W. Biederman
2008-07-10 0:23 ` Jeremy Fitzhardinge
2008-07-09 20:35 ` H. Peter Anvin
2008-07-09 20:39 ` Arjan van de Ven
2008-07-09 20:44 ` H. Peter Anvin
2008-07-09 20:50 ` Jeremy Fitzhardinge
2008-07-09 21:12 ` H. Peter Anvin
2008-07-09 21:26 ` Jeremy Fitzhardinge
2008-07-09 21:37 ` H. Peter Anvin
2008-07-09 22:10 ` Eric W. Biederman
2008-07-09 22:23 ` H. Peter Anvin
2008-07-09 23:54 ` Eric W. Biederman
2008-07-10 16:22 ` Mike Travis
2008-07-10 16:25 ` H. Peter Anvin
2008-07-10 16:35 ` Christoph Lameter
2008-07-10 16:39 ` H. Peter Anvin
2008-07-10 16:47 ` Christoph Lameter
2008-07-10 17:21 ` Jeremy Fitzhardinge
2008-07-10 17:31 ` Christoph Lameter
2008-07-10 17:48 ` Jeremy Fitzhardinge
2008-07-10 18:00 ` H. Peter Anvin
2008-07-10 17:20 ` Mike Travis
2008-07-10 17:07 ` Jeremy Fitzhardinge
2008-07-10 17:12 ` Christoph Lameter
2008-07-10 17:25 ` Jeremy Fitzhardinge
2008-07-10 17:34 ` Christoph Lameter
2008-07-10 17:41 ` Mike Travis
2008-07-10 18:01 ` H. Peter Anvin
2008-07-10 20:51 ` Christoph Lameter
2008-07-10 20:58 ` H. Peter Anvin
2008-07-10 21:07 ` Christoph Lameter
2008-07-10 21:11 ` H. Peter Anvin
2008-07-11 15:32 ` Christoph Lameter
2008-07-11 16:07 ` H. Peter Anvin
2008-07-11 16:57 ` Eric W. Biederman
2008-07-11 17:10 ` H. Peter Anvin
2008-07-10 21:26 ` Eric W. Biederman
2008-07-10 18:48 ` Eric W. Biederman
2008-07-10 18:54 ` Jeremy Fitzhardinge
2008-07-10 19:18 ` Eric W. Biederman
2008-07-10 19:56 ` Jeremy Fitzhardinge
2008-07-10 20:22 ` Eric W. Biederman
2008-07-10 20:54 ` Jeremy Fitzhardinge
2008-07-11 6:59 ` Rusty Russell
2008-07-10 20:25 ` Eric W. Biederman
2008-07-10 17:57 ` H. Peter Anvin
2008-07-10 18:08 ` H. Peter Anvin
2008-07-09 20:46 ` Jeremy Fitzhardinge
2008-07-09 20:14 ` Arjan van de Ven
2008-07-09 20:33 ` Eric W. Biederman
2008-07-09 21:01 ` Ingo Molnar
2008-07-09 21:39 ` Mike Travis
2008-07-09 21:47 ` Jeremy Fitzhardinge
2008-07-09 21:55 ` Eric W. Biederman
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=48769D1E.3030003@sgi.com \
--to=travis@sgi.com \
--cc=akpm@linux-foundation.org \
--cc=arjan@infradead.org \
--cc=cl@linux-foundation.org \
--cc=ebiederm@xmission.com \
--cc=hpa@zytor.com \
--cc=jeremy@goop.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=steiner@sgi.com \
/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.