The x86 architecture supports additional memory ordering flags to mark lock critical sections for hardware lock elision. These must be specified in addition to an existing memory order to atomic intrinsics.
__ATOMIC_HLE_ACQUIRE
__ATOMIC_ACQUIRE
or stronger. __ATOMIC_HLE_RELEASE
__ATOMIC_RELEASE
or stronger. When a lock acquire fails, it is required for good performance to abort the transaction quickly. This can be done with a _mm_pause
.
#include <immintrin.h> // For _mm_pause int lockvar; /* Acquire lock with lock elision */ while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE)) _mm_pause(); /* Abort failed transaction */ ... /* Free lock with lock elision */ __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
© Free Software Foundation
Licensed under the GNU Free Documentation License, Version 1.3.
https://gcc.gnu.org/onlinedocs/gcc-6.3.0/gcc/x86-specific-memory-model-extensions-for-transactional-memory.html