This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Unify pthread_spin_[try]lock implementations.
- From: Carlos O'Donell <carlos_odonell at mentor dot com>
- To: Maxim Kuvyrkov <maxim at codesourcery dot com>
- Cc: Roland McGrath <roland at hack dot frob dot com>, Andrew Haley <aph at redhat dot com>,David Miller <davem at davemloft dot net>, "Joseph S. Myers"<joseph at codesourcery dot com>, Richard Sandiford <rdsandiford at googlemail dot com>,<libc-ports at sourceware dot org>, GLIBC Devel <libc-alpha at sourceware dot org>, ChrisMetcalf <cmetcalf at tilera dot com>
- Date: Wed, 15 Aug 2012 12:42:54 -0400
- Subject: Re: [PATCH] Unify pthread_spin_[try]lock implementations.
- References: <Pine.LNX.4.64.1206282306320.20312@digraph.polyomino.org.uk> <65B470D2-4D01-4BA1-AEC5-A72C0006EA22@codesourcery.com> <20120711081441.73BB22C093@topped-with-meat.com> <20120711.012509.1325789838255235021.davem@davemloft.net> <4FFD3CD9.4030206@redhat.com> <84304C03-6A49-4263-9016-05486EDC0E98@codesourcery.com> <4FFD4114.9000806@redhat.com> <E1DB09C1-0E3E-4088-9793-C0CAB80B5084@codesourcery.com> <20120711112235.B28CA2C099@topped-with-meat.com> <7FBB4F87-9FF3-4239-818F-5A38C8094011@codesourcery.com> <20120725181300.DD1812C0B5@topped-with-meat.com> <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