public inbox for libc-ports@sourceware.org
 help / color / mirror / Atom feed
From: Carlos O'Donell <carlos_odonell@mentor.com>
To: Maxim Kuvyrkov <maxim@codesourcery.com>
Cc: Roland McGrath <roland@hack.frob.com>,
	Andrew Haley <aph@redhat.com>,	David Miller <davem@davemloft.net>,
	"Joseph S. Myers"	<joseph@codesourcery.com>,
	Richard Sandiford <rdsandiford@googlemail.com>,
	<libc-ports@sourceware.org>,
	GLIBC Devel <libc-alpha@sourceware.org>,
	Chris Metcalf <cmetcalf@tilera.com>
Subject: Re: [PATCH] Unify pthread_spin_[try]lock implementations.
Date: Wed, 15 Aug 2012 16:43:00 -0000	[thread overview]
Message-ID: <502BD18E.2000802@mentor.com> (raw)
In-Reply-To: <36A2FFD8-0C98-4AB6-8C64-2EEC5CC67A63@codesourcery.com>

On 8/14/2012 11:16 PM, Maxim Kuvyrkov wrote:
> On 26/07/2012, at 6:13 AM, Roland McGrath wrote:
> 
>> Here I think the reasonable thing to do is:
>>
>> /* A machine-specific version can define SPIN_LOCK_READS_BETWEEN_CMPXCHG
>>   to the number of plain reads that it's optimal to spin on between uses
>>   of atomic_compare_and_exchange_val_acq.  If spinning forever is optimal
>>   then use -1.  If no plain reads here would ever be optimal, use 0.  */
>> #ifndef SPIN_LOCK_READS_BETWEEN_CMPXCHG
>> # warning machine-dependent file should define SPIN_LOCK_READS_BETWEEN_CMPXCHG
>> # define SPIN_LOCK_READS_BETWEEN_CMPXCHG 1000
>> #endif
>>
>> Then ARM et al can do:
>>
>> /* Machine-dependent rationale about the selection of this value.  */
>> #define SPIN_LOCK_READS_BETWEEN_CMPXCHG 1000
>> #include <nptl/pthread_spin_lock.c>
>>
>> while Teil will use -1.
>>
>>> +      if (PTHREAD_SPIN_LOCK_WAIT)
>>
>> Don't use implicit boolean coercion.
>> Use "if (SPIN_LOCK_READS_BETWEEN_CMPXCHG >= 0)".
>>
>>> +	{
>>> +	  int wait = PTHREAD_SPIN_LOCK_WAIT;
>>> +
>>> +	  while (*lock != 0 && --wait)
>>> +	    ;
>>
>> Write it:
>> 	while (wait > 0 && *lock != 0)
>> 	  --wait;
>>
>> That handles the SPIN_LOCK_READS_BETWEEN_CMPXCHG==0 case implicitly,
>> avoids the ugly empty statement, and doesn't use implicit coercion.
> 
> OK, the updated patch attached.  Is this what you had in mind?
> 
> As before the patch touches ARM, HPPA, M68K and MIPS.
> 
> The patch also adds another optimization to use atomic_exchange_acq for the first attempt at acquiring the lock.  For a free lock atomic_exchange is, generally, faster; while atomic_compare_and_exchange is better for waiting on a contended lock.  Same rationale applies to pthread_spin_trylock.
> 
> This patch builds on MIPS and regression testing is in progress.  OK to apply is tests are fine?
> 
> Thank you,
> 
> --
> Maxim Kuvyrkov
> CodeSourcery / Mentor Graphics
> 
> 
> Add generic versions of pthread_spin_lock and pthread_spin_trylock.
> 
> 	2012-07-09  Maxim Kuvyrkov  <maxim@codesourcery.com>
> 
> 	* nptl/pthread_spin_lock.c: New file.
> 	* nptl/pthread_spin_trylock.c: New file.
> 
> 	ports/
> 	* sysdeps/arm/nptl/pthread_spin_lock.c: Use generic code.
> 	* sysdeps/arm/nptl/pthread_spin_trylock.c: Remove, use generic version.
> 
> 	* sysdeps/hppa/nptl/pthread_spin_lock.c: Use generic code.
> 	* sysdeps/hppa/nptl/pthread_spin_trylock.c: Remove, use generic version.
> 
> 	* sysdeps/m68k/nptl/pthread_spin_lock.c: Use generic code.
> 	* sysdeps/m68k/nptl/pthread_spin_trylock.c: Remove, use generic version.
> 
> 	* sysdeps/mips/nptl/pthread_spin_lock.S: Remove, use generic version.
> 	* sysdeps/mips/nptl/pthread_spin_lock.c: New file.
> 	* sysdeps/mips/nptl/pthread_spin_trylock.S: Remove, use generic version.
> ---
>  nptl/pthread_spin_lock.c                       |   69 ++++++++++++++++++++++++
>  nptl/pthread_spin_trylock.c                    |   27 +++++++++
>  ports/sysdeps/arm/nptl/pthread_spin_lock.c     |   16 +-----
>  ports/sysdeps/arm/nptl/pthread_spin_trylock.c  |   26 ---------
>  ports/sysdeps/hppa/nptl/pthread_spin_lock.c    |   24 +-------
>  ports/sysdeps/hppa/nptl/pthread_spin_trylock.c |   33 -----------
>  ports/sysdeps/m68k/nptl/pthread_spin_lock.c    |   16 +-----
>  ports/sysdeps/m68k/nptl/pthread_spin_trylock.c |   27 ---------
>  ports/sysdeps/mips/nptl/pthread_spin_lock.S    |   36 ------------
>  ports/sysdeps/mips/nptl/pthread_spin_lock.c    |   19 +++++++
>  ports/sysdeps/mips/nptl/pthread_spin_trylock.S |   40 --------------
>  11 files changed, 124 insertions(+), 209 deletions(-)
>  create mode 100644 nptl/pthread_spin_lock.c
>  create mode 100644 nptl/pthread_spin_trylock.c
>  delete mode 100644 ports/sysdeps/arm/nptl/pthread_spin_trylock.c
>  delete mode 100644 ports/sysdeps/hppa/nptl/pthread_spin_trylock.c
>  delete mode 100644 ports/sysdeps/m68k/nptl/pthread_spin_trylock.c
>  delete mode 100644 ports/sysdeps/mips/nptl/pthread_spin_lock.S
>  create mode 100644 ports/sysdeps/mips/nptl/pthread_spin_lock.c
>  delete mode 100644 ports/sysdeps/mips/nptl/pthread_spin_trylock.S
> 
> diff --git a/ports/sysdeps/hppa/nptl/pthread_spin_lock.c b/ports/sysdeps/hppa/nptl/pthread_spin_lock.c
> index bcf2240..7c86768 100644
> --- a/ports/sysdeps/hppa/nptl/pthread_spin_lock.c
> +++ b/ports/sysdeps/hppa/nptl/pthread_spin_lock.c
> @@ -1,4 +1,4 @@
> -/* Copyright (C) 2005 Free Software Foundation, Inc.
> +/* Copyright (C) 2005-2012 Free Software Foundation, Inc.
>     This file is part of the GNU C Library.
>  
>     The GNU C Library is free software; you can redistribute it and/or
> @@ -15,23 +15,5 @@
>     License along with the GNU C Library.  If not, see
>     <http://www.gnu.org/licenses/>.  */
>  
> -#include <atomic.h>
> -#include "pthreadP.h"
> -
> -int
> -pthread_spin_lock (pthread_spinlock_t *lock)
> -{
> -#if 0
> -  volatile unsigned int *addr = __ldcw_align (lock);
> -
> -  while (__ldcw (addr) == 0)
> -    while (*addr == 0) ;
> -
> -  return 0;
> -#endif
> -
> -  while (atomic_compare_and_exchange_val_acq(lock, 1, 0) == 1)
> -    while (*lock == 1);
> -  
> -  return 0;
> -}
> +#define SPIN_LOCK_READS_BETWEEN_CMPXCHG 1000
> +#include_next <nptl/pthread_spin_lock.c>

OK for hppa.

> diff --git a/ports/sysdeps/hppa/nptl/pthread_spin_trylock.c b/ports/sysdeps/hppa/nptl/pthread_spin_trylock.c
> deleted file mode 100644
> index a802861..0000000
> --- a/ports/sysdeps/hppa/nptl/pthread_spin_trylock.c
> +++ /dev/null
> @@ -1,33 +0,0 @@
> -/* Copyright (C) 2005 Free Software Foundation, Inc.
> -   This file is part of the GNU C Library.
> -
> -   The GNU C Library is free software; you can redistribute it and/or
> -   modify it under the terms of the GNU Lesser General Public
> -   License as published by the Free Software Foundation; either
> -   version 2.1 of the License, or (at your option) any later version.
> -
> -   The GNU C Library is distributed in the hope that it will be useful,
> -   but WITHOUT ANY WARRANTY; without even the implied warranty of
> -   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> -   Lesser General Public License for more details.
> -
> -   You should have received a copy of the GNU Lesser General Public
> -   License along with the GNU C Library.  If not, see
> -   <http://www.gnu.org/licenses/>.  */
> -
> -#include <errno.h>
> -#include <atomic.h>
> -#include "pthreadP.h"
> -
> -int
> -pthread_spin_trylock (pthread_spinlock_t *lock)
> -{
> -#if 0
> -  volatile unsigned int *a = __ldcw_align (lock);
> -
> -  return __ldcw (a) ? 0 : EBUSY;
> -#endif
> -
> -  return atomic_compare_and_exchange_val_acq(lock, 1, 0) ? EBUSY : 0;
> -
> -}

OK for hppa.

Cheers,
Carlos.
-- 
Carlos O'Donell
Mentor Graphics / CodeSourcery
carlos_odonell@mentor.com
carlos@codesourcery.com
+1 (613) 963 1026

  parent reply	other threads:[~2012-08-15 16:43 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-06-14  4:36 [PATCH 2/3, MIPS] Rewrite MIPS' pthread_spin_[try]lock using __atomic_* builtins Maxim Kuvyrkov
2012-06-28 23:10 ` Joseph S. Myers
2012-07-11  5:58   ` [PATCH] Unify pthread_spin_[try]lock implementations Maxim Kuvyrkov
2012-07-11  8:15     ` Roland McGrath
2012-07-11  8:25       ` David Miller
2012-07-11  8:44         ` Andrew Haley
2012-07-11  8:54           ` Maxim Kuvyrkov
2012-07-11  9:02             ` Andrew Haley
2012-07-11  9:05               ` Maxim Kuvyrkov
2012-07-11 11:22                 ` Roland McGrath
2012-07-11 14:52                   ` Chris Metcalf
2012-07-11 22:16                   ` Maxim Kuvyrkov
2012-07-25 18:13                     ` Roland McGrath
2012-07-25 19:04                       ` Chris Metcalf
2012-07-25 20:22                         ` Roland McGrath
2012-07-25 20:29                           ` Chris Metcalf
2012-07-25 20:43                             ` Roland McGrath
2012-08-15  3:17                       ` Maxim Kuvyrkov
2012-08-15 16:27                         ` Roland McGrath
2012-08-15 16:39                           ` Maxim Kuvyrkov
2012-08-15 16:44                             ` Roland McGrath
2012-08-15 16:53                               ` Maxim Kuvyrkov
2012-08-15 16:56                                 ` Roland McGrath
2012-08-15 17:05                                   ` Maxim Kuvyrkov
2012-08-15 17:05                             ` Jeff Law
2012-08-15 17:11                               ` Roland McGrath
2012-08-15 17:24                                 ` Jeff Law
2012-08-15 16:40                         ` Joseph S. Myers
2012-08-15 16:43                         ` Carlos O'Donell [this message]
2012-08-15 22:00                         ` Andreas Schwab
2012-08-16 10:21                         ` Torvald Riegel
2012-08-16 12:40                           ` Chris Metcalf
2012-08-22 23:16                           ` Maxim Kuvyrkov
2012-07-11  8:26       ` Maxim Kuvyrkov

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=502BD18E.2000802@mentor.com \
    --to=carlos_odonell@mentor.com \
    --cc=aph@redhat.com \
    --cc=cmetcalf@tilera.com \
    --cc=davem@davemloft.net \
    --cc=joseph@codesourcery.com \
    --cc=libc-alpha@sourceware.org \
    --cc=libc-ports@sourceware.org \
    --cc=maxim@codesourcery.com \
    --cc=rdsandiford@googlemail.com \
    --cc=roland@hack.frob.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 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 read-only IMAP folder(s) and NNTP newsgroup(s).