From: Nick Piggin <npiggin@suse.de>
To: Linus Torvalds <torvalds@osdl.org>
Cc: Nick Piggin <npiggin@suse.de>,
Linux Memory Management <linux-mm@kvack.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Hugh Dickins <hugh@veritas.com>, Andrew Morton <akpm@osdl.org>,
Andrea Arcangeli <andrea@suse.de>,
David Miller <davem@davemloft.net>
Subject: Re: [patch 0/4] mm: de-skew page refcount
Date: Wed, 18 Jan 2006 18:05:58 +0100 [thread overview]
Message-ID: <20060118170558.GE28418@wotan.suse.de> (raw)
In-Reply-To: <Pine.LNX.4.64.0601180830520.3240@g5.osdl.org>
On Wed, Jan 18, 2006 at 08:38:44AM -0800, Linus Torvalds wrote:
>
>
> On Wed, 18 Jan 2006, Nick Piggin wrote:
> >
> > The following patchset (against 2.6.16-rc1 + migrate race fixes) uses the new
> > atomic ops to do away with the offset page refcounting, and simplify the race
> > that it was designed to cover.
> >
> > This allows some nice optimisations
>
> Why?
>
> The real downside is that "atomic_inc_nonzero()" is a lot more expensive
> than checking for zero on x86 (and x86-64).
>
> The reason it's offset is that on architectures that automatically test
> the _result_ of an atomic op (ie x86[-64]), it's easy to see when
> something _becomes_ negative or _becomes_ zero, and that's what
>
> atomic_add_negative
> atomic_inc_and_test
>
> are optimized for (there's also "atomic_dec_and_test()" which reacts on
> the count becoming zero, but that doesn't have a pairing: there's no way
> to react to the count becoming one for the increment operation, so the
> "atomic_dec_and_test()" is used for things where zero means "free it").
>
> Nothing else can be done that fast on x86. Everything else requires an
> insane "load, update, cmpxchg" sequence.
>
Yes, I realise inc_not_zero isn't as fast as dec_and_test on x86(-64).
In this case when the cacheline will already be exclusive I bet it isn't
that much of a difference (in the noise from my testing).
> So I disagree with this patch series. It has real downsides. There's a
> reason we have the offset.
>
Yes, there is a reason, I detailed it in the changelog and got rid of it.
> "nice optimizations"
They're in this patchset.
Nick
next prev parent reply other threads:[~2006-01-18 17:06 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-01-18 10:40 [patch 0/4] mm: de-skew page refcount Nick Piggin
2006-01-18 10:40 ` [patch 1/4] mm: page refcount use atomic primitives Nick Piggin
2006-01-18 10:40 ` [patch 2/4] mm: PageLRU no testset Nick Piggin
2006-01-19 17:48 ` Nikita Danilov
2006-01-19 18:10 ` Nick Piggin
2006-01-18 10:40 ` [patch 3/3] mm: PageActive " Nick Piggin
2006-01-18 14:13 ` Marcelo Tosatti
2006-01-19 14:50 ` Nick Piggin
2006-01-19 16:52 ` Marcelo Tosatti
2006-01-19 20:02 ` Nick Piggin
2006-01-19 21:41 ` Marcelo Tosatti
2006-01-18 10:41 ` [patch 4/4] mm: less atomic ops Nick Piggin
2006-01-18 16:38 ` [patch 0/4] mm: de-skew page refcount Linus Torvalds
2006-01-18 17:05 ` Nick Piggin [this message]
2006-01-18 19:27 ` Linus Torvalds
2006-01-19 14:00 ` Nick Piggin
2006-01-19 16:36 ` Linus Torvalds
2006-01-19 17:06 ` Nick Piggin
2006-01-19 17:27 ` Linus Torvalds
2006-01-19 17:38 ` Nick Piggin
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=20060118170558.GE28418@wotan.suse.de \
--to=npiggin@suse.de \
--cc=akpm@osdl.org \
--cc=andrea@suse.de \
--cc=davem@davemloft.net \
--cc=hugh@veritas.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=torvalds@osdl.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 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).