glibc/sysdeps/alpha/atomic-machine.h
Stefan Liebler 12d2dd7060 Optimize generic spinlock code and use C11 like atomic macros.
This patch optimizes the generic spinlock code.

The type pthread_spinlock_t is a typedef to volatile int on all archs.
Passing a volatile pointer to the atomic macros which are not mapped to the
C11 atomic builtins can lead to extra stores and loads to stack if such
a macro creates a temporary variable by using "__typeof (*(mem)) tmp;".
Thus, those macros which are used by spinlock code - atomic_exchange_acquire,
atomic_load_relaxed, atomic_compare_exchange_weak - have to be adjusted.
According to the comment from  Szabolcs Nagy, the type of a cast expression is
unqualified (see http://www.open-std.org/jtc1/sc22/wg14/www/docs/dr_423.htm):
__typeof ((__typeof (*(mem)) *(mem)) tmp;
Thus from spinlock perspective the variable tmp is of type int instead of
type volatile int.  This patch adjusts those macros in include/atomic.h.
With this construct GCC >= 5 omits the extra stores and loads.

The atomic macros are replaced by the C11 like atomic macros and thus
the code is aligned to it.  The pthread_spin_unlock implementation is now
using release memory order instead of sequentially consistent memory order.
The issue with passed volatile int pointers applies to the C11 like atomic
macros as well as the ones used before.

I've added a glibc_likely hint to the first atomic exchange in
pthread_spin_lock in order to return immediately to the caller if the lock is
free.  Without the hint, there is an additional jump if the lock is free.

I've added the atomic_spin_nop macro within the loop of plain reads.
The plain reads are also realized by C11 like atomic_load_relaxed macro.

The new define ATOMIC_EXCHANGE_USES_CAS determines if the first try to acquire
the spinlock in pthread_spin_lock or pthread_spin_trylock is an exchange
or a CAS.  This is defined in atomic-machine.h for all architectures.

The define SPIN_LOCK_READS_BETWEEN_CMPXCHG is now removed.
There is no technical reason for throwing in a CAS every now and then,
and so far we have no evidence that it can improve performance.
If that would be the case, we have to adjust other spin-waiting loops
elsewhere, too!  Using a CAS loop without plain reads is not a good idea
on many targets and wasn't used by one.  Thus there is now no option to
do so.

Architectures are now using the generic spinlock automatically if they
do not provide an own implementation.  Thus the pthread_spin_lock.c files
in sysdeps folder are deleted.

ChangeLog:

	* NEWS: Mention new spinlock implementation.
	* include/atomic.h:
	(__atomic_val_bysize): Cast type to omit volatile qualifier.
	(atomic_exchange_acq): Likewise.
	(atomic_load_relaxed): Likewise.
	(ATOMIC_EXCHANGE_USES_CAS): Check definition.
	* nptl/pthread_spin_init.c (pthread_spin_init):
	Use atomic_store_relaxed.
	* nptl/pthread_spin_lock.c (pthread_spin_lock):
	Use C11-like atomic macros.
	* nptl/pthread_spin_trylock.c (pthread_spin_trylock):
	Likewise.
	* nptl/pthread_spin_unlock.c (pthread_spin_unlock):
	Use atomic_store_release.
	* sysdeps/aarch64/nptl/pthread_spin_lock.c: Delete File.
	* sysdeps/arm/nptl/pthread_spin_lock.c: Likewise.
	* sysdeps/hppa/nptl/pthread_spin_lock.c: Likewise.
	* sysdeps/m68k/nptl/pthread_spin_lock.c: Likewise.
	* sysdeps/microblaze/nptl/pthread_spin_lock.c: Likewise.
	* sysdeps/mips/nptl/pthread_spin_lock.c: Likewise.
	* sysdeps/nios2/nptl/pthread_spin_lock.c: Likewise.
	* sysdeps/aarch64/atomic-machine.h (ATOMIC_EXCHANGE_USES_CAS): Define.
	* sysdeps/alpha/atomic-machine.h: Likewise.
	* sysdeps/arm/atomic-machine.h: Likewise.
	* sysdeps/i386/atomic-machine.h: Likewise.
	* sysdeps/ia64/atomic-machine.h: Likewise.
	* sysdeps/m68k/coldfire/atomic-machine.h: Likewise.
	* sysdeps/m68k/m680x0/m68020/atomic-machine.h: Likewise.
	* sysdeps/microblaze/atomic-machine.h: Likewise.
	* sysdeps/mips/atomic-machine.h: Likewise.
	* sysdeps/powerpc/powerpc32/atomic-machine.h: Likewise.
	* sysdeps/powerpc/powerpc64/atomic-machine.h: Likewise.
	* sysdeps/s390/atomic-machine.h: Likewise.
	* sysdeps/sparc/sparc32/atomic-machine.h: Likewise.
	* sysdeps/sparc/sparc32/sparcv9/atomic-machine.h: Likewise.
	* sysdeps/sparc/sparc64/atomic-machine.h: Likewise.
	* sysdeps/tile/tilegx/atomic-machine.h: Likewise.
	* sysdeps/tile/tilepro/atomic-machine.h: Likewise.
	* sysdeps/unix/sysv/linux/hppa/atomic-machine.h: Likewise.
	* sysdeps/unix/sysv/linux/m68k/coldfire/atomic-machine.h: Likewise.
	* sysdeps/unix/sysv/linux/nios2/atomic-machine.h: Likewise.
	* sysdeps/unix/sysv/linux/sh/atomic-machine.h: Likewise.
	* sysdeps/x86_64/atomic-machine.h: Likewise.
2017-06-06 09:41:56 +02:00

371 lines
12 KiB
C

/* Copyright (C) 2003-2017 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 <stdint.h>
typedef int8_t atomic8_t;
typedef uint8_t uatomic8_t;
typedef int_fast8_t atomic_fast8_t;
typedef uint_fast8_t uatomic_fast8_t;
typedef int16_t atomic16_t;
typedef uint16_t uatomic16_t;
typedef int_fast16_t atomic_fast16_t;
typedef uint_fast16_t uatomic_fast16_t;
typedef int32_t atomic32_t;
typedef uint32_t uatomic32_t;
typedef int_fast32_t atomic_fast32_t;
typedef uint_fast32_t uatomic_fast32_t;
typedef int64_t atomic64_t;
typedef uint64_t uatomic64_t;
typedef int_fast64_t atomic_fast64_t;
typedef uint_fast64_t uatomic_fast64_t;
typedef intptr_t atomicptr_t;
typedef uintptr_t uatomicptr_t;
typedef intmax_t atomic_max_t;
typedef uintmax_t uatomic_max_t;
#define __HAVE_64B_ATOMICS 1
#define USE_ATOMIC_COMPILER_BUILTINS 0
/* XXX Is this actually correct? */
#define ATOMIC_EXCHANGE_USES_CAS 1
#ifdef UP
# define __MB /* nothing */
#else
# define __MB " mb\n"
#endif
/* Compare and exchange. For all of the "xxx" routines, we expect a
"__prev" and a "__cmp" variable to be provided by the enclosing scope,
in which values are returned. */
#define __arch_compare_and_exchange_xxx_8_int(mem, new, old, mb1, mb2) \
({ \
unsigned long __tmp, __snew, __addr64; \
__asm__ __volatile__ ( \
mb1 \
" andnot %[__addr8],7,%[__addr64]\n" \
" insbl %[__new],%[__addr8],%[__snew]\n" \
"1: ldq_l %[__tmp],0(%[__addr64])\n" \
" extbl %[__tmp],%[__addr8],%[__prev]\n" \
" cmpeq %[__prev],%[__old],%[__cmp]\n" \
" beq %[__cmp],2f\n" \
" mskbl %[__tmp],%[__addr8],%[__tmp]\n" \
" or %[__snew],%[__tmp],%[__tmp]\n" \
" stq_c %[__tmp],0(%[__addr64])\n" \
" beq %[__tmp],1b\n" \
mb2 \
"2:" \
: [__prev] "=&r" (__prev), \
[__snew] "=&r" (__snew), \
[__tmp] "=&r" (__tmp), \
[__cmp] "=&r" (__cmp), \
[__addr64] "=&r" (__addr64) \
: [__addr8] "r" (mem), \
[__old] "Ir" ((uint64_t)(uint8_t)(uint64_t)(old)), \
[__new] "r" (new) \
: "memory"); \
})
#define __arch_compare_and_exchange_xxx_16_int(mem, new, old, mb1, mb2) \
({ \
unsigned long __tmp, __snew, __addr64; \
__asm__ __volatile__ ( \
mb1 \
" andnot %[__addr16],7,%[__addr64]\n" \
" inswl %[__new],%[__addr16],%[__snew]\n" \
"1: ldq_l %[__tmp],0(%[__addr64])\n" \
" extwl %[__tmp],%[__addr16],%[__prev]\n" \
" cmpeq %[__prev],%[__old],%[__cmp]\n" \
" beq %[__cmp],2f\n" \
" mskwl %[__tmp],%[__addr16],%[__tmp]\n" \
" or %[__snew],%[__tmp],%[__tmp]\n" \
" stq_c %[__tmp],0(%[__addr64])\n" \
" beq %[__tmp],1b\n" \
mb2 \
"2:" \
: [__prev] "=&r" (__prev), \
[__snew] "=&r" (__snew), \
[__tmp] "=&r" (__tmp), \
[__cmp] "=&r" (__cmp), \
[__addr64] "=&r" (__addr64) \
: [__addr16] "r" (mem), \
[__old] "Ir" ((uint64_t)(uint16_t)(uint64_t)(old)), \
[__new] "r" (new) \
: "memory"); \
})
#define __arch_compare_and_exchange_xxx_32_int(mem, new, old, mb1, mb2) \
({ \
__asm__ __volatile__ ( \
mb1 \
"1: ldl_l %[__prev],%[__mem]\n" \
" cmpeq %[__prev],%[__old],%[__cmp]\n" \
" beq %[__cmp],2f\n" \
" mov %[__new],%[__cmp]\n" \
" stl_c %[__cmp],%[__mem]\n" \
" beq %[__cmp],1b\n" \
mb2 \
"2:" \
: [__prev] "=&r" (__prev), \
[__cmp] "=&r" (__cmp) \
: [__mem] "m" (*(mem)), \
[__old] "Ir" ((uint64_t)(atomic32_t)(uint64_t)(old)), \
[__new] "Ir" (new) \
: "memory"); \
})
#define __arch_compare_and_exchange_xxx_64_int(mem, new, old, mb1, mb2) \
({ \
__asm__ __volatile__ ( \
mb1 \
"1: ldq_l %[__prev],%[__mem]\n" \
" cmpeq %[__prev],%[__old],%[__cmp]\n" \
" beq %[__cmp],2f\n" \
" mov %[__new],%[__cmp]\n" \
" stq_c %[__cmp],%[__mem]\n" \
" beq %[__cmp],1b\n" \
mb2 \
"2:" \
: [__prev] "=&r" (__prev), \
[__cmp] "=&r" (__cmp) \
: [__mem] "m" (*(mem)), \
[__old] "Ir" ((uint64_t)(old)), \
[__new] "Ir" (new) \
: "memory"); \
})
/* For all "bool" routines, we return FALSE if exchange succesful. */
#define __arch_compare_and_exchange_bool_8_int(mem, new, old, mb1, mb2) \
({ unsigned long __prev; int __cmp; \
__arch_compare_and_exchange_xxx_8_int(mem, new, old, mb1, mb2); \
!__cmp; })
#define __arch_compare_and_exchange_bool_16_int(mem, new, old, mb1, mb2) \
({ unsigned long __prev; int __cmp; \
__arch_compare_and_exchange_xxx_16_int(mem, new, old, mb1, mb2); \
!__cmp; })
#define __arch_compare_and_exchange_bool_32_int(mem, new, old, mb1, mb2) \
({ unsigned long __prev; int __cmp; \
__arch_compare_and_exchange_xxx_32_int(mem, new, old, mb1, mb2); \
!__cmp; })
#define __arch_compare_and_exchange_bool_64_int(mem, new, old, mb1, mb2) \
({ unsigned long __prev; int __cmp; \
__arch_compare_and_exchange_xxx_64_int(mem, new, old, mb1, mb2); \
!__cmp; })
/* For all "val" routines, return the old value whether exchange
successful or not. */
#define __arch_compare_and_exchange_val_8_int(mem, new, old, mb1, mb2) \
({ unsigned long __prev; int __cmp; \
__arch_compare_and_exchange_xxx_8_int(mem, new, old, mb1, mb2); \
(typeof (*mem))__prev; })
#define __arch_compare_and_exchange_val_16_int(mem, new, old, mb1, mb2) \
({ unsigned long __prev; int __cmp; \
__arch_compare_and_exchange_xxx_16_int(mem, new, old, mb1, mb2); \
(typeof (*mem))__prev; })
#define __arch_compare_and_exchange_val_32_int(mem, new, old, mb1, mb2) \
({ unsigned long __prev; int __cmp; \
__arch_compare_and_exchange_xxx_32_int(mem, new, old, mb1, mb2); \
(typeof (*mem))__prev; })
#define __arch_compare_and_exchange_val_64_int(mem, new, old, mb1, mb2) \
({ unsigned long __prev; int __cmp; \
__arch_compare_and_exchange_xxx_64_int(mem, new, old, mb1, mb2); \
(typeof (*mem))__prev; })
/* Compare and exchange with "acquire" semantics, ie barrier after. */
#define atomic_compare_and_exchange_bool_acq(mem, new, old) \
__atomic_bool_bysize (__arch_compare_and_exchange_bool, int, \
mem, new, old, "", __MB)
#define atomic_compare_and_exchange_val_acq(mem, new, old) \
__atomic_val_bysize (__arch_compare_and_exchange_val, int, \
mem, new, old, "", __MB)
/* Compare and exchange with "release" semantics, ie barrier before. */
#define atomic_compare_and_exchange_val_rel(mem, new, old) \
__atomic_val_bysize (__arch_compare_and_exchange_val, int, \
mem, new, old, __MB, "")
/* Atomically store value and return the previous value. */
#define __arch_exchange_8_int(mem, value, mb1, mb2) \
({ \
unsigned long __tmp, __addr64, __sval; __typeof(*mem) __ret; \
__asm__ __volatile__ ( \
mb1 \
" andnot %[__addr8],7,%[__addr64]\n" \
" insbl %[__value],%[__addr8],%[__sval]\n" \
"1: ldq_l %[__tmp],0(%[__addr64])\n" \
" extbl %[__tmp],%[__addr8],%[__ret]\n" \
" mskbl %[__tmp],%[__addr8],%[__tmp]\n" \
" or %[__sval],%[__tmp],%[__tmp]\n" \
" stq_c %[__tmp],0(%[__addr64])\n" \
" beq %[__tmp],1b\n" \
mb2 \
: [__ret] "=&r" (__ret), \
[__sval] "=&r" (__sval), \
[__tmp] "=&r" (__tmp), \
[__addr64] "=&r" (__addr64) \
: [__addr8] "r" (mem), \
[__value] "r" (value) \
: "memory"); \
__ret; })
#define __arch_exchange_16_int(mem, value, mb1, mb2) \
({ \
unsigned long __tmp, __addr64, __sval; __typeof(*mem) __ret; \
__asm__ __volatile__ ( \
mb1 \
" andnot %[__addr16],7,%[__addr64]\n" \
" inswl %[__value],%[__addr16],%[__sval]\n" \
"1: ldq_l %[__tmp],0(%[__addr64])\n" \
" extwl %[__tmp],%[__addr16],%[__ret]\n" \
" mskwl %[__tmp],%[__addr16],%[__tmp]\n" \
" or %[__sval],%[__tmp],%[__tmp]\n" \
" stq_c %[__tmp],0(%[__addr64])\n" \
" beq %[__tmp],1b\n" \
mb2 \
: [__ret] "=&r" (__ret), \
[__sval] "=&r" (__sval), \
[__tmp] "=&r" (__tmp), \
[__addr64] "=&r" (__addr64) \
: [__addr16] "r" (mem), \
[__value] "r" (value) \
: "memory"); \
__ret; })
#define __arch_exchange_32_int(mem, value, mb1, mb2) \
({ \
signed int __tmp; __typeof(*mem) __ret; \
__asm__ __volatile__ ( \
mb1 \
"1: ldl_l %[__ret],%[__mem]\n" \
" mov %[__val],%[__tmp]\n" \
" stl_c %[__tmp],%[__mem]\n" \
" beq %[__tmp],1b\n" \
mb2 \
: [__ret] "=&r" (__ret), \
[__tmp] "=&r" (__tmp) \
: [__mem] "m" (*(mem)), \
[__val] "Ir" (value) \
: "memory"); \
__ret; })
#define __arch_exchange_64_int(mem, value, mb1, mb2) \
({ \
unsigned long __tmp; __typeof(*mem) __ret; \
__asm__ __volatile__ ( \
mb1 \
"1: ldq_l %[__ret],%[__mem]\n" \
" mov %[__val],%[__tmp]\n" \
" stq_c %[__tmp],%[__mem]\n" \
" beq %[__tmp],1b\n" \
mb2 \
: [__ret] "=&r" (__ret), \
[__tmp] "=&r" (__tmp) \
: [__mem] "m" (*(mem)), \
[__val] "Ir" (value) \
: "memory"); \
__ret; })
#define atomic_exchange_acq(mem, value) \
__atomic_val_bysize (__arch_exchange, int, mem, value, "", __MB)
#define atomic_exchange_rel(mem, value) \
__atomic_val_bysize (__arch_exchange, int, mem, value, __MB, "")
/* Atomically add value and return the previous (unincremented) value. */
#define __arch_exchange_and_add_8_int(mem, value, mb1, mb2) \
({ __builtin_trap (); 0; })
#define __arch_exchange_and_add_16_int(mem, value, mb1, mb2) \
({ __builtin_trap (); 0; })
#define __arch_exchange_and_add_32_int(mem, value, mb1, mb2) \
({ \
signed int __tmp; __typeof(*mem) __ret; \
__asm__ __volatile__ ( \
mb1 \
"1: ldl_l %[__ret],%[__mem]\n" \
" addl %[__ret],%[__val],%[__tmp]\n" \
" stl_c %[__tmp],%[__mem]\n" \
" beq %[__tmp],1b\n" \
mb2 \
: [__ret] "=&r" (__ret), \
[__tmp] "=&r" (__tmp) \
: [__mem] "m" (*(mem)), \
[__val] "Ir" ((signed int)(value)) \
: "memory"); \
__ret; })
#define __arch_exchange_and_add_64_int(mem, value, mb1, mb2) \
({ \
unsigned long __tmp; __typeof(*mem) __ret; \
__asm__ __volatile__ ( \
mb1 \
"1: ldq_l %[__ret],%[__mem]\n" \
" addq %[__ret],%[__val],%[__tmp]\n" \
" stq_c %[__tmp],%[__mem]\n" \
" beq %[__tmp],1b\n" \
mb2 \
: [__ret] "=&r" (__ret), \
[__tmp] "=&r" (__tmp) \
: [__mem] "m" (*(mem)), \
[__val] "Ir" ((unsigned long)(value)) \
: "memory"); \
__ret; })
/* ??? Barrier semantics for atomic_exchange_and_add appear to be
undefined. Use full barrier for now, as that's safe. */
#define atomic_exchange_and_add(mem, value) \
__atomic_val_bysize (__arch_exchange_and_add, int, mem, value, __MB, __MB)
/* ??? Blah, I'm lazy. Implement these later. Can do better than the
compare-and-exchange loop provided by generic code.
#define atomic_decrement_if_positive(mem)
#define atomic_bit_test_set(mem, bit)
*/
#ifndef UP
# define atomic_full_barrier() __asm ("mb" : : : "memory");
# define atomic_read_barrier() __asm ("mb" : : : "memory");
# define atomic_write_barrier() __asm ("wmb" : : : "memory");
#endif