[PATCH 3/5] x86: Define atomic_full_barrier using __sync_synchronize

[Adhemerval Zanella Netto]

On 08/09/25 12:33, Uros Bizjak wrote:
> On Mon, Sep 8, 2025 at 2:38 PM Uros Bizjak <ubizjak@gmail.com> wrote:
>>
>> For x86_64 targets, __sync_synchronize emits a full 64-bit
>> 'LOCK ORQ $0x0,(%rsp)' instead of 'LOCK ORL $0x0,(%rsp)'.
>>
>> No functional changes.
>>
>> Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
>> Cc: Florian Weimer <fweimer@redhat.com>
>> Cc: Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>
>> Cc: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
>> Cc: Collin Funk <collin.funk1@gmail.com>
>> Cc: H.J.Lu <hjl.tools@gmail.com>
>> Cc: Carlos O'Donell <carlos@redhat.com>
>> ---
>>  sysdeps/x86/atomic-machine.h | 8 ++------
>>  1 file changed, 2 insertions(+), 6 deletions(-)
>>
>> diff --git a/sysdeps/x86/atomic-machine.h b/sysdeps/x86/atomic-machine.h
>> index d5b2d49031..c0c2c3437a 100644
>> --- a/sysdeps/x86/atomic-machine.h
>> +++ b/sysdeps/x86/atomic-machine.h
>> @@ -26,15 +26,14 @@
>>
>>  #ifdef __x86_64__
>>  # define __HAVE_64B_ATOMICS            1
>> -# define SP_REG                                "rsp"
>>  #else
>>  /* Since the Pentium, i386 CPUs have supported 64-bit atomics, but the
>>     i386 psABI supplement provides only 4-byte alignment for uint64_t
>>     inside structs, so it is currently not possible to use 64-bit
>>     atomics on this platform.  */
>>  # define __HAVE_64B_ATOMICS            0
>> -# define SP_REG                                "esp"
>>  #endif
>> +
>>  #define ATOMIC_EXCHANGE_USES_CAS       0
>>
>>  #define atomic_compare_and_exchange_val_acq(mem, newval, oldval) \
>> @@ -74,10 +73,7 @@
>>  #define catomic_exchange_and_add(mem, value) \
>>    __atomic_fetch_add (mem, value, __ATOMIC_ACQUIRE)
>>
>> -/* We don't use mfence because it is supposedly slower due to having to
>> -   provide stronger guarantees (e.g., regarding self-modifying code).  */
>> -#define atomic_full_barrier() \
>> -    __asm __volatile (LOCK_PREFIX "orl $0, (%%" SP_REG ")" ::: "memory")
>> +#define atomic_full_barrier() __sync_synchronize ()
> 
> Oh... this could be:
> 
> #define atomic_full_barrier() __atomic_thread_fence (__ATOMIC_SEQ_CST);
> 
> to keep primitives consistent with c11. This one is functionally
> equivalent to __sync_synchronize ().

Either way works for me.

LGTM, thanks.

Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>