All of lore.kernel.org
 help / color / mirror / Atom feed
From: Diego Calleja <diegocg@gmail.com>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Andi Kleen <andi@firstfloor.org>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: [PATCH prototype] [0/8] Predictive bitmaps for ELF executables
Date: Thu, 20 Mar 2008 01:15:20 +0100	[thread overview]
Message-ID: <20080320011520.60e151be.diegocg@gmail.com> (raw)
In-Reply-To: <20080319020440.80379d50.akpm@linux-foundation.org>

El Wed, 19 Mar 2008 02:04:40 -0700, Andrew Morton <akpm@linux-foundation.org> escribió:

> Assuming that all users have the same access pattern might be inefficient,
> a little bit.  There might be some advantage to making it per-user, dunno.

In the Dark Side of operating systems, the prefetching system they use
can log several access patterns for a single executable, because a single
executable can have different behaviours even for the same user, depending
on what parameters the executable is passed and what COM machinery it
uses. For example, wmplayer.exe can play a dvd, rip a CD, listen to a music
stream, etc...diferent usages, different access patterns. Linux probably faces
the same problem (bash, cat...)

A alternative design for a userspace solution that doesn't needs LD_PRELOAD
is to use CONFIG_PROC_EVENTS to get notifications of what processes are
started, which can be used to poll its /proc files or try to preload data
(asynchronously, and a bit hacky maybe).

But if a kernel patch is really needed to implement this properly, maybe
it'd be worth to take a look at the prefetch project that the Ubuntu guys
are apparently going to merge in the next ubuntu development release (8.10)...
https://wiki.ubuntu.com/DesktopTeam/Specs/Prefetch

There are even kernel patches:
http://code.google.com/p/prefetch/source/browse/tags/soc2007-end/trunk/kernel-patches/2.6.22/submitted/0001-prefetch-core.diff
http://code.google.com/p/prefetch/source/browse/tags/soc2007-end/trunk/kernel-patches/2.6.22/submitted/0002-prefetch-boot.diff

WARNING: multiple messages have this Message-ID (diff)
From: Diego Calleja <diegocg@gmail.com>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Andi Kleen <andi@firstfloor.org>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: [PATCH prototype] [0/8] Predictive bitmaps for ELF executables
Date: Thu, 20 Mar 2008 01:15:20 +0100	[thread overview]
Message-ID: <20080320011520.60e151be.diegocg@gmail.com> (raw)
In-Reply-To: <20080319020440.80379d50.akpm@linux-foundation.org>

El Wed, 19 Mar 2008 02:04:40 -0700, Andrew Morton <akpm@linux-foundation.org> escribio:

> Assuming that all users have the same access pattern might be inefficient,
> a little bit.  There might be some advantage to making it per-user, dunno.

In the Dark Side of operating systems, the prefetching system they use
can log several access patterns for a single executable, because a single
executable can have different behaviours even for the same user, depending
on what parameters the executable is passed and what COM machinery it
uses. For example, wmplayer.exe can play a dvd, rip a CD, listen to a music
stream, etc...diferent usages, different access patterns. Linux probably faces
the same problem (bash, cat...)

A alternative design for a userspace solution that doesn't needs LD_PRELOAD
is to use CONFIG_PROC_EVENTS to get notifications of what processes are
started, which can be used to poll its /proc files or try to preload data
(asynchronously, and a bit hacky maybe).

But if a kernel patch is really needed to implement this properly, maybe
it'd be worth to take a look at the prefetch project that the Ubuntu guys
are apparently going to merge in the next ubuntu development release (8.10)...
https://wiki.ubuntu.com/DesktopTeam/Specs/Prefetch

There are even kernel patches:
http://code.google.com/p/prefetch/source/browse/tags/soc2007-end/trunk/kernel-patches/2.6.22/submitted/0001-prefetch-core.diff
http://code.google.com/p/prefetch/source/browse/tags/soc2007-end/trunk/kernel-patches/2.6.22/submitted/0002-prefetch-boot.diff

--
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:"dont@kvack.org"> email@kvack.org </a>

  parent reply	other threads:[~2008-03-20  0:37 UTC|newest]

Thread overview: 82+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-03-18  1:09 [PATCH prototype] [0/8] Predictive bitmaps for ELF executables Andi Kleen
2008-03-18  1:09 ` Andi Kleen
2008-03-18  1:09 ` [PATCH prototype] [1/8] Give ELF shdr types a name Andi Kleen
2008-03-18  1:09   ` Andi Kleen
2008-03-18  1:09 ` [PATCH prototype] [2/8] Add support to override mmap exec write protection with O_FORCEWRITE Andi Kleen
2008-03-18  1:09   ` Andi Kleen
2008-03-18  1:09 ` [PATCH prototype] [3/8] Make readahead max pinned value a sysctl Andi Kleen
2008-03-18  1:09   ` Andi Kleen
2008-03-18  1:09 ` [PATCH prototype] [4/8] Add readahead function to read-ahead based on a bitmap Andi Kleen
2008-03-18  1:09   ` Andi Kleen
2008-03-18  1:09 ` [PATCH prototype] [5/8] Add ELF constants for pbitmaps Andi Kleen
2008-03-18  1:09   ` Andi Kleen
2008-03-18  1:09 ` [PATCH prototype] [6/8] Core predictive bitmap engine Andi Kleen
2008-03-18  1:09   ` Andi Kleen
2008-03-18  1:09 ` [PATCH prototype] [7/8] Add the sysctls to control pbitmaps Andi Kleen
2008-03-18  1:09   ` Andi Kleen
2008-03-18  1:09 ` [PATCH prototype] [8/8] Add mmap_full_slurp support Andi Kleen
2008-03-18  1:09   ` Andi Kleen
2008-03-18  7:36 ` [PATCH prototype] [0/8] Predictive bitmaps for ELF executables Andrew Morton
2008-03-18  7:36   ` Andrew Morton
2008-03-18 14:18   ` Andi Kleen
2008-03-18 14:18     ` Andi Kleen
2008-03-18 16:57     ` Andrew Morton
2008-03-18 16:57       ` Andrew Morton
2008-03-18 17:20       ` Andi Kleen
2008-03-18 17:20         ` Andi Kleen
2008-03-18 17:44         ` Andrew Morton
2008-03-18 17:44           ` Andrew Morton
2008-03-19  8:32           ` Andi Kleen
2008-03-19  8:32             ` Andi Kleen
2008-03-19  9:04             ` Andrew Morton
2008-03-19  9:04               ` Andrew Morton
2008-03-19 22:45               ` Ulrich Drepper
2008-03-19 22:45                 ` Ulrich Drepper
2008-03-19 23:12                 ` Andrew Morton
2008-03-19 23:12                   ` Andrew Morton
2008-03-20  0:09                   ` David Miller
2008-03-20  0:09                     ` David Miller, Andrew Morton
2008-03-20  9:00                 ` Andi Kleen
2008-03-20  9:00                   ` Andi Kleen
2008-03-21 17:15                   ` Ulrich Drepper
2008-03-21 17:15                     ` Ulrich Drepper
2008-03-21 17:26                     ` Andi Kleen
2008-03-21 17:26                       ` Andi Kleen
2008-03-22  4:36                       ` Ulrich Drepper
2008-03-22  4:36                         ` Ulrich Drepper
2008-03-22  7:17                         ` Andi Kleen
2008-03-22  7:17                           ` Andi Kleen
2008-03-22  7:24                           ` Nicholas Miell
2008-03-22  7:24                             ` Nicholas Miell
2008-03-22  9:10                             ` Andi Kleen
2008-03-22  9:10                               ` Andi Kleen
2008-03-22 10:16                               ` Nicholas Miell
2008-03-22 10:16                                 ` Nicholas Miell
2008-03-22 14:29                                 ` Andi Kleen
2008-03-22 14:29                                   ` Andi Kleen
2008-03-23 13:25                                   ` Pavel Machek
2008-03-23 13:25                                     ` Pavel Machek
2008-03-23 17:08                                     ` Andi Kleen
2008-03-23 17:08                                       ` Andi Kleen
2008-03-24 16:24                                       ` Pavel Machek
2008-03-24 16:24                                         ` Pavel Machek
2008-03-24  4:20                               ` Ulrich Drepper
2008-03-24  4:20                                 ` Ulrich Drepper
2008-03-24  5:16                                 ` Nicholas Miell
2008-03-24  5:16                                   ` Nicholas Miell
2008-03-24  5:26                                   ` Andi Kleen
2008-03-24  5:26                                     ` Andi Kleen
2008-03-24 19:42                                   ` Ulrich Drepper
2008-03-24 19:42                                     ` Ulrich Drepper
2008-03-24 21:47                                     ` Nicholas Miell
2008-03-24 21:47                                       ` Nicholas Miell
2008-03-25  7:54                                     ` Andi Kleen
2008-03-25  7:54                                       ` Andi Kleen
2008-03-26 18:15                                       ` Ulrich Drepper
2008-03-26 18:15                                         ` Ulrich Drepper
2008-03-26 18:54                                         ` Andi Kleen
2008-03-26 18:54                                           ` Andi Kleen
2008-03-22  4:38                       ` Ulrich Drepper
2008-03-22  4:38                         ` Ulrich Drepper
2008-03-20  0:15               ` Diego Calleja [this message]
2008-03-20  0:15                 ` Diego Calleja

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=20080320011520.60e151be.diegocg@gmail.com \
    --to=diegocg@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=andi@firstfloor.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.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 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.