#ifndef GOOGLE_PROTOBUF_ATOMICOPS_INTERNALS_X86_GCC_H_
#define GOOGLE_PROTOBUF_ATOMICOPS_INTERNALS_X86_GCC_H_
namespace google {
namespace protobuf {
namespace internal {
struct AtomicOps_x86CPUFeatureStruct {
bool has_amd_lock_mb_bug;
bool has_sse2;
};
extern struct AtomicOps_x86CPUFeatureStruct AtomicOps_Internalx86CPUFeatures;
#define ATOMICOPS_COMPILER_BARRIER() __asm__ __volatile__("" : : : "memory")
inline Atomic32 NoBarrier_CompareAndSwap(volatile Atomic32* ptr,
Atomic32 old_value,
Atomic32 new_value) {
Atomic32 prev;
__asm__ __volatile__("lock; cmpxchgl %1,%2"
: "=a" (prev)
: "q" (new_value), "m" (*ptr), "0" (old_value)
: "memory");
return prev;
}
inline Atomic32 NoBarrier_AtomicExchange(volatile Atomic32* ptr,
Atomic32 new_value) {
__asm__ __volatile__("xchgl %1,%0"
: "=r" (new_value)
: "m" (*ptr), "0" (new_value)
: "memory");
return new_value;
}
inline Atomic32 NoBarrier_AtomicIncrement(volatile Atomic32* ptr,
Atomic32 increment) {
Atomic32 temp = increment;
__asm__ __volatile__("lock; xaddl %0,%1"
: "+r" (temp), "+m" (*ptr)
: : "memory");
return temp + increment;
}
inline Atomic32 Barrier_AtomicIncrement(volatile Atomic32* ptr,
Atomic32 increment) {
Atomic32 temp = increment;
__asm__ __volatile__("lock; xaddl %0,%1"
: "+r" (temp), "+m" (*ptr)
: : "memory");
if (AtomicOps_Internalx86CPUFeatures.has_amd_lock_mb_bug) {
__asm__ __volatile__("lfence" : : : "memory");
}
return temp + increment;
}
inline Atomic32 Acquire_CompareAndSwap(volatile Atomic32* ptr,
Atomic32 old_value,
Atomic32 new_value) {
Atomic32 x = NoBarrier_CompareAndSwap(ptr, old_value, new_value);
if (AtomicOps_Internalx86CPUFeatures.has_amd_lock_mb_bug) {
__asm__ __volatile__("lfence" : : : "memory");
}
return x;
}
inline Atomic32 Release_CompareAndSwap(volatile Atomic32* ptr,
Atomic32 old_value,
Atomic32 new_value) {
return NoBarrier_CompareAndSwap(ptr, old_value, new_value);
}
inline void NoBarrier_Store(volatile Atomic32* ptr, Atomic32 value) {
*ptr = value;
}
#if defined(__x86_64__)
inline void MemoryBarrier() {
__asm__ __volatile__("mfence" : : : "memory");
}
inline void Acquire_Store(volatile Atomic32* ptr, Atomic32 value) {
*ptr = value;
MemoryBarrier();
}
#else
inline void MemoryBarrier() {
if (AtomicOps_Internalx86CPUFeatures.has_sse2) {
__asm__ __volatile__("mfence" : : : "memory");
} else {
Atomic32 x = 0;
NoBarrier_AtomicExchange(&x, 0);
}
}
inline void Acquire_Store(volatile Atomic32* ptr, Atomic32 value) {
if (AtomicOps_Internalx86CPUFeatures.has_sse2) {
*ptr = value;
__asm__ __volatile__("mfence" : : : "memory");
} else {
NoBarrier_AtomicExchange(ptr, value);
}
}
#endif
inline void Release_Store(volatile Atomic32* ptr, Atomic32 value) {
ATOMICOPS_COMPILER_BARRIER();
*ptr = value;
}
inline Atomic32 NoBarrier_Load(volatile const Atomic32* ptr) {
return *ptr;
}
inline Atomic32 Acquire_Load(volatile const Atomic32* ptr) {
Atomic32 value = *ptr;
ATOMICOPS_COMPILER_BARRIER();
return value;
}
inline Atomic32 Release_Load(volatile const Atomic32* ptr) {
MemoryBarrier();
return *ptr;
}
#if defined(__x86_64__)
inline Atomic64 NoBarrier_CompareAndSwap(volatile Atomic64* ptr,
Atomic64 old_value,
Atomic64 new_value) {
Atomic64 prev;
__asm__ __volatile__("lock; cmpxchgq %1,%2"
: "=a" (prev)
: "q" (new_value), "m" (*ptr), "0" (old_value)
: "memory");
return prev;
}
inline Atomic64 NoBarrier_AtomicExchange(volatile Atomic64* ptr,
Atomic64 new_value) {
__asm__ __volatile__("xchgq %1,%0"
: "=r" (new_value)
: "m" (*ptr), "0" (new_value)
: "memory");
return new_value;
}
inline Atomic64 NoBarrier_AtomicIncrement(volatile Atomic64* ptr,
Atomic64 increment) {
Atomic64 temp = increment;
__asm__ __volatile__("lock; xaddq %0,%1"
: "+r" (temp), "+m" (*ptr)
: : "memory");
return temp + increment;
}
inline Atomic64 Barrier_AtomicIncrement(volatile Atomic64* ptr,
Atomic64 increment) {
Atomic64 temp = increment;
__asm__ __volatile__("lock; xaddq %0,%1"
: "+r" (temp), "+m" (*ptr)
: : "memory");
if (AtomicOps_Internalx86CPUFeatures.has_amd_lock_mb_bug) {
__asm__ __volatile__("lfence" : : : "memory");
}
return temp + increment;
}
inline void NoBarrier_Store(volatile Atomic64* ptr, Atomic64 value) {
*ptr = value;
}
inline void Acquire_Store(volatile Atomic64* ptr, Atomic64 value) {
*ptr = value;
MemoryBarrier();
}
inline void Release_Store(volatile Atomic64* ptr, Atomic64 value) {
ATOMICOPS_COMPILER_BARRIER();
*ptr = value;
}
inline Atomic64 NoBarrier_Load(volatile const Atomic64* ptr) {
return *ptr;
}
inline Atomic64 Acquire_Load(volatile const Atomic64* ptr) {
Atomic64 value = *ptr;
ATOMICOPS_COMPILER_BARRIER();
return value;
}
inline Atomic64 Release_Load(volatile const Atomic64* ptr) {
MemoryBarrier();
return *ptr;
}
inline Atomic64 Acquire_CompareAndSwap(volatile Atomic64* ptr,
Atomic64 old_value,
Atomic64 new_value) {
Atomic64 x = NoBarrier_CompareAndSwap(ptr, old_value, new_value);
if (AtomicOps_Internalx86CPUFeatures.has_amd_lock_mb_bug) {
__asm__ __volatile__("lfence" : : : "memory");
}
return x;
}
inline Atomic64 Release_CompareAndSwap(volatile Atomic64* ptr,
Atomic64 old_value,
Atomic64 new_value) {
return NoBarrier_CompareAndSwap(ptr, old_value, new_value);
}
#endif
}
}
}
#undef ATOMICOPS_COMPILER_BARRIER
#endif