From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Will Deacon <will.deacon@arm.com>, Arnd Bergmann <arnd@arndb.de>,
"Figo. zhang" <figo1802@gmail.com>,
Aswin Chandramouleeswaran <aswin@hp.com>,
Rik van Riel <riel@redhat.com>, Waiman Long <waiman.long@hp.com>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>,
"linux-arch@vger.kernel.org" <linux-arch@vger.kernel.org>,
Andi Kleen <andi@firstfloor.org>,
George Spelvin <linux@horizon.com>,
Tim Chen <tim.c.chen@linux.intel.com>,
Michel Lespinasse <walken@google.com>,
Ingo Molnar <mingo@elte.hu>,
Peter Hurley <peter@hurleysoftware.com>,
"H. Peter Anvin" <hpa@zytor.com>,
Andrew Morton <akpm@linux-foundation.org>,
linux-mm <linux-mm@kvack.org>, Alex Shi <alex.shi@linaro.org>,
Andrea Arcangeli <aarcange@redhat.com>
Subject: Re: [PATCH v6 4/5] MCS Lock: Barrier corrections
Date: Sat, 30 Nov 2013 09:38:43 -0800 [thread overview]
Message-ID: <20131130173843.GZ4137@linux.vnet.ibm.com> (raw)
In-Reply-To: <CA+55aFwHgnH4h0YwybThQjvicFCVbGbwaAy3Fw0b738gJMtqBA@mail.gmail.com>
On Fri, Nov 29, 2013 at 08:44:41AM -0800, Linus Torvalds wrote:
> On Nov 29, 2013 8:18 AM, "Will Deacon" <will.deacon@arm.com> wrote:
> >
> > To get some sort of
> > idea, I tried adding a dmb to the start of spin_unlock on ARMv7 and I saw
> a
> > 3% performance hit in hackbench on my dual-cluster board.
>
> Don't do a dmb. Just do a dummy release. You just said that on arm64 a
> unlock+lock is a memory barrier, so just make the mb__before_spinlock() be
> a dummy store with release to the stack..
>
> That should be noticeably cheaper than a full dmb.
Cute! I like it! ;-)
Thanx, Paul
--
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>
WARNING: multiple messages have this Message-ID (diff)
From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Will Deacon <will.deacon@arm.com>, Arnd Bergmann <arnd@arndb.de>,
"Figo. zhang" <figo1802@gmail.com>,
Aswin Chandramouleeswaran <aswin@hp.com>,
Rik van Riel <riel@redhat.com>, Waiman Long <waiman.long@hp.com>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>,
"linux-arch@vger.kernel.org" <linux-arch@vger.kernel.org>,
Andi Kleen <andi@firstfloor.org>,
George Spelvin <linux@horizon.com>,
Tim Chen <tim.c.chen@linux.intel.com>,
Michel Lespinasse <walken@google.com>,
Ingo Molnar <mingo@elte.hu>,
Peter Hurley <peter@hurleysoftware.com>,
"H. Peter Anvin" <hpa@zytor.com>,
Andrew Morton <akpm@linux-foundation.org>,
linux-mm <linux-mm@kvack.org>, Alex Shi <alex.shi@linaro.org>,
Andrea Arcangeli <aarcange@redhat.com>,
Scott J Norton <scott.norton@hp.com>,
Thomas Gleixner <tglx@linutronix.de>,
Dave Hansen <dave.hansen@intel.com>,
Peter Zijlstra <peterz@infradead.org>,
Matthew R Wilcox <matthew.r.wilcox@intel.com>,
Davidlohr Bueso <davidlohr.bueso@hp.com>
Subject: Re: [PATCH v6 4/5] MCS Lock: Barrier corrections
Date: Sat, 30 Nov 2013 09:38:43 -0800 [thread overview]
Message-ID: <20131130173843.GZ4137@linux.vnet.ibm.com> (raw)
In-Reply-To: <CA+55aFwHgnH4h0YwybThQjvicFCVbGbwaAy3Fw0b738gJMtqBA@mail.gmail.com>
On Fri, Nov 29, 2013 at 08:44:41AM -0800, Linus Torvalds wrote:
> On Nov 29, 2013 8:18 AM, "Will Deacon" <will.deacon@arm.com> wrote:
> >
> > To get some sort of
> > idea, I tried adding a dmb to the start of spin_unlock on ARMv7 and I saw
> a
> > 3% performance hit in hackbench on my dual-cluster board.
>
> Don't do a dmb. Just do a dummy release. You just said that on arm64 a
> unlock+lock is a memory barrier, so just make the mb__before_spinlock() be
> a dummy store with release to the stack..
>
> That should be noticeably cheaper than a full dmb.
Cute! I like it! ;-)
Thanx, Paul
--
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>
next prev parent reply other threads:[~2013-11-30 17:38 UTC|newest]
Thread overview: 239+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <cover.1384885312.git.tim.c.chen@linux.intel.com>
2013-11-20 1:37 ` [PATCH v6 0/5] MCS Lock: MCS lock code cleanup and optimizations Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 10:19 ` Will Deacon
2013-11-20 10:19 ` Will Deacon
2013-11-20 12:50 ` Paul E. McKenney
2013-11-20 12:50 ` Paul E. McKenney
2013-11-20 17:00 ` Will Deacon
2013-11-20 17:00 ` Will Deacon
2013-11-20 17:14 ` Paul E. McKenney
2013-11-20 17:14 ` Paul E. McKenney
2013-11-20 17:00 ` Tim Chen
2013-11-20 17:00 ` Tim Chen
2013-11-20 17:16 ` Paul E. McKenney
2013-11-20 17:16 ` Paul E. McKenney
2013-11-20 1:37 ` [PATCH v6 1/5] MCS Lock: Restructure the MCS lock defines and locking code into its own file Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` [PATCH v6 2/5] MCS Lock: optimizations and extra comments Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` [PATCH v6 3/5] MCS Lock: Move mcs_lock/unlock function into its own file Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` [PATCH v6 4/5] MCS Lock: Barrier corrections Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 15:31 ` Paul E. McKenney
2013-11-20 15:31 ` Paul E. McKenney
2013-11-20 15:31 ` Paul E. McKenney
2013-11-20 15:46 ` Will Deacon
2013-11-20 15:46 ` Will Deacon
2013-11-20 17:14 ` Paul E. McKenney
2013-11-20 17:14 ` Paul E. McKenney
2013-11-20 18:43 ` Tim Chen
2013-11-20 18:43 ` Tim Chen
2013-11-20 19:06 ` Paul E. McKenney
2013-11-20 19:06 ` Paul E. McKenney
2013-11-20 20:36 ` Tim Chen
2013-11-20 20:36 ` Tim Chen
2013-11-20 21:44 ` Paul E. McKenney
2013-11-20 21:44 ` Paul E. McKenney
2013-11-20 23:51 ` Tim Chen
2013-11-20 23:51 ` Tim Chen
2013-11-21 4:53 ` Paul E. McKenney
2013-11-21 4:53 ` Paul E. McKenney
2013-11-21 10:17 ` Will Deacon
2013-11-21 10:17 ` Will Deacon
2013-11-21 13:16 ` Paul E. McKenney
2013-11-21 13:16 ` Paul E. McKenney
2013-11-21 10:45 ` Peter Zijlstra
2013-11-21 10:45 ` Peter Zijlstra
2013-11-21 13:18 ` Paul E. McKenney
2013-11-21 13:18 ` Paul E. McKenney
2013-11-21 22:27 ` Linus Torvalds
2013-11-21 22:27 ` Linus Torvalds
2013-11-21 22:52 ` Paul E. McKenney
2013-11-21 22:52 ` Paul E. McKenney
2013-11-22 0:09 ` Linus Torvalds
2013-11-22 0:09 ` Linus Torvalds
2013-11-22 4:08 ` Paul E. McKenney
2013-11-22 4:08 ` Paul E. McKenney
2013-11-22 4:25 ` Linus Torvalds
2013-11-22 4:25 ` Linus Torvalds
2013-11-22 6:23 ` Paul E. McKenney
2013-11-22 6:23 ` Paul E. McKenney
2013-11-22 15:16 ` Ingo Molnar
2013-11-22 15:16 ` Ingo Molnar
2013-11-22 18:49 ` Paul E. McKenney
2013-11-22 18:49 ` Paul E. McKenney
2013-11-22 19:06 ` Linus Torvalds
2013-11-22 19:06 ` Linus Torvalds
2013-11-22 20:06 ` Paul E. McKenney
2013-11-22 20:06 ` Paul E. McKenney
2013-11-22 20:09 ` Linus Torvalds
2013-11-22 20:09 ` Linus Torvalds
2013-11-22 20:37 ` Paul E. McKenney
2013-11-22 20:37 ` Paul E. McKenney
2013-11-22 21:01 ` Linus Torvalds
2013-11-22 21:01 ` Linus Torvalds
2013-11-22 21:52 ` Paul E. McKenney
2013-11-22 21:52 ` Paul E. McKenney
2013-11-22 22:19 ` Linus Torvalds
2013-11-22 22:19 ` Linus Torvalds
2013-11-23 0:25 ` Paul E. McKenney
2013-11-23 0:25 ` Paul E. McKenney
2013-11-23 0:42 ` Linus Torvalds
2013-11-23 0:42 ` Linus Torvalds
2013-11-23 1:36 ` Paul E. McKenney
2013-11-23 1:36 ` Paul E. McKenney
2013-11-23 2:11 ` Linus Torvalds
2013-11-23 2:11 ` Linus Torvalds
2013-11-23 4:05 ` Paul E. McKenney
2013-11-23 4:05 ` Paul E. McKenney
2013-11-23 11:24 ` Ingo Molnar
2013-11-23 11:24 ` Ingo Molnar
2013-11-23 17:06 ` Paul E. McKenney
2013-11-23 17:06 ` Paul E. McKenney
2013-11-26 12:02 ` Ingo Molnar
2013-11-26 12:02 ` Ingo Molnar
2013-11-26 19:28 ` Paul E. McKenney
2013-11-26 19:28 ` Paul E. McKenney
2013-11-23 20:21 ` Linus Torvalds
2013-11-23 20:21 ` Linus Torvalds
2013-11-23 20:39 ` Linus Torvalds
2013-11-23 20:39 ` Linus Torvalds
2013-11-25 12:09 ` Peter Zijlstra
2013-11-25 12:09 ` Peter Zijlstra
2013-11-25 17:18 ` Will Deacon
2013-11-25 17:18 ` Will Deacon
2013-11-25 17:56 ` Paul E. McKenney
2013-11-25 17:56 ` Paul E. McKenney
2013-11-25 17:54 ` Paul E. McKenney
2013-11-25 17:54 ` Paul E. McKenney
2013-11-23 21:29 ` Peter Zijlstra
2013-11-23 21:29 ` Peter Zijlstra
2013-11-23 22:24 ` Linus Torvalds
2013-11-23 22:24 ` Linus Torvalds
2013-11-25 17:53 ` Paul E. McKenney
2013-11-25 17:53 ` Paul E. McKenney
2013-11-25 18:21 ` Peter Zijlstra
2013-11-25 18:21 ` Peter Zijlstra
2013-11-21 11:03 ` Peter Zijlstra
2013-11-21 11:03 ` Peter Zijlstra
2013-11-21 12:56 ` Peter Zijlstra
2013-11-21 12:56 ` Peter Zijlstra
2013-11-21 13:20 ` Paul E. McKenney
2013-11-21 13:20 ` Paul E. McKenney
2013-11-21 17:25 ` Paul E. McKenney
2013-11-21 17:25 ` Paul E. McKenney
2013-11-21 21:52 ` Peter Zijlstra
2013-11-21 21:52 ` Peter Zijlstra
2013-11-21 22:18 ` Paul E. McKenney
2013-11-21 22:18 ` Paul E. McKenney
2013-11-22 15:58 ` Peter Zijlstra
2013-11-22 15:58 ` Peter Zijlstra
2013-11-22 18:26 ` Paul E. McKenney
2013-11-22 18:26 ` Paul E. McKenney
2013-11-22 18:51 ` Peter Zijlstra
2013-11-22 18:51 ` Peter Zijlstra
2013-11-22 18:59 ` Paul E. McKenney
2013-11-22 18:59 ` Paul E. McKenney
2013-11-25 17:35 ` Peter Zijlstra
2013-11-25 17:35 ` Peter Zijlstra
2013-11-25 18:02 ` Paul E. McKenney
2013-11-25 18:02 ` Paul E. McKenney
2013-11-25 18:24 ` Peter Zijlstra
2013-11-25 18:24 ` Peter Zijlstra
2013-11-25 18:34 ` Tim Chen
2013-11-25 18:34 ` Tim Chen
2013-11-25 18:27 ` Peter Zijlstra
2013-11-25 18:27 ` Peter Zijlstra
2013-11-25 23:52 ` Paul E. McKenney
2013-11-25 23:52 ` Paul E. McKenney
2013-11-26 9:59 ` Peter Zijlstra
2013-11-26 9:59 ` Peter Zijlstra
2013-11-26 17:11 ` Paul E. McKenney
2013-11-26 17:11 ` Paul E. McKenney
2013-11-26 17:18 ` Peter Zijlstra
2013-11-26 17:18 ` Peter Zijlstra
2013-11-26 19:00 ` Linus Torvalds
2013-11-26 19:00 ` Linus Torvalds
2013-11-26 19:20 ` Paul E. McKenney
2013-11-26 19:20 ` Paul E. McKenney
2013-11-26 19:32 ` Linus Torvalds
2013-11-26 19:32 ` Linus Torvalds
2013-11-26 22:51 ` Paul E. McKenney
2013-11-26 22:51 ` Paul E. McKenney
2013-11-26 23:58 ` Linus Torvalds
2013-11-26 23:58 ` Linus Torvalds
2013-11-27 0:21 ` Thomas Gleixner
2013-11-27 0:21 ` Thomas Gleixner
2013-11-27 0:39 ` Paul E. McKenney
2013-11-27 0:39 ` Paul E. McKenney
2013-11-27 1:05 ` Linus Torvalds
2013-11-27 1:05 ` Linus Torvalds
2013-11-27 1:31 ` Paul E. McKenney
2013-11-27 1:31 ` Paul E. McKenney
2013-11-27 10:16 ` Will Deacon
2013-11-27 10:16 ` Will Deacon
2013-11-27 17:11 ` Paul E. McKenney
2013-11-27 17:11 ` Paul E. McKenney
2013-11-28 11:40 ` Will Deacon
2013-11-28 11:40 ` Will Deacon
2013-11-28 17:38 ` Paul E. McKenney
2013-11-28 17:38 ` Paul E. McKenney
2013-11-28 18:03 ` Will Deacon
2013-11-28 18:03 ` Will Deacon
2013-11-28 18:27 ` Paul E. McKenney
2013-11-28 18:27 ` Paul E. McKenney
2013-11-28 18:53 ` Will Deacon
2013-11-28 18:53 ` Will Deacon
2013-11-28 19:50 ` Paul E. McKenney
2013-11-28 19:50 ` Paul E. McKenney
2013-11-29 16:17 ` Will Deacon
2013-11-29 16:17 ` Will Deacon
2013-11-29 16:44 ` Linus Torvalds
2013-11-29 16:44 ` Linus Torvalds
2013-11-29 18:18 ` Will Deacon
2013-11-29 18:18 ` Will Deacon
2013-11-30 17:38 ` Paul E. McKenney [this message]
2013-11-30 17:38 ` Paul E. McKenney
2013-11-26 19:21 ` Peter Zijlstra
2013-11-26 19:21 ` Peter Zijlstra
2013-11-27 16:58 ` Oleg Nesterov
2013-11-27 16:58 ` Oleg Nesterov
2013-11-26 23:08 ` Benjamin Herrenschmidt
2013-11-26 23:08 ` Benjamin Herrenschmidt
2013-11-25 23:55 ` H. Peter Anvin
2013-11-25 23:55 ` H. Peter Anvin
2013-11-26 3:16 ` Paul E. McKenney
2013-11-26 3:16 ` Paul E. McKenney
2013-11-27 0:46 ` H. Peter Anvin
2013-11-27 0:46 ` H. Peter Anvin
2013-11-27 1:07 ` Linus Torvalds
2013-11-27 1:07 ` Linus Torvalds
2013-11-27 1:27 ` Paul E. McKenney
2013-11-27 1:27 ` Paul E. McKenney
2013-11-27 2:59 ` H. Peter Anvin
2013-11-27 2:59 ` H. Peter Anvin
2013-11-25 18:52 ` H. Peter Anvin
2013-11-25 18:52 ` H. Peter Anvin
2013-11-25 22:58 ` Tim Chen
2013-11-25 22:58 ` Tim Chen
2013-11-25 23:28 ` H. Peter Anvin
2013-11-25 23:28 ` H. Peter Anvin
2013-11-25 23:51 ` Paul E. McKenney
2013-11-25 23:51 ` Paul E. McKenney
2013-11-25 23:36 ` Paul E. McKenney
2013-11-25 23:36 ` Paul E. McKenney
2013-12-04 21:26 ` Andi Kleen
2013-12-04 21:26 ` Andi Kleen
2013-12-04 22:07 ` Paul E. McKenney
2013-12-04 22:07 ` Paul E. McKenney
2013-11-21 13:19 ` Paul E. McKenney
2013-11-21 13:19 ` Paul E. McKenney
2013-11-20 1:37 ` [PATCH v6 5/5] MCS Lock: Allows for architecture specific mcs lock and unlock Tim Chen
2013-11-20 1:37 ` Tim Chen
2013-11-20 1:37 ` Tim Chen
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=20131130173843.GZ4137@linux.vnet.ibm.com \
--to=paulmck@linux.vnet.ibm.com \
--cc=aarcange@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=alex.shi@linaro.org \
--cc=andi@firstfloor.org \
--cc=arnd@arndb.de \
--cc=aswin@hp.com \
--cc=figo1802@gmail.com \
--cc=hpa@zytor.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux@horizon.com \
--cc=mingo@elte.hu \
--cc=peter@hurleysoftware.com \
--cc=raghavendra.kt@linux.vnet.ibm.com \
--cc=riel@redhat.com \
--cc=tim.c.chen@linux.intel.com \
--cc=torvalds@linux-foundation.org \
--cc=waiman.long@hp.com \
--cc=walken@google.com \
--cc=will.deacon@arm.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.