std::atomic_thread_fence (3) - Linux Manuals

std::atomic_thread_fence: std::atomic_thread_fence

NAME

std::atomic_thread_fence - std::atomic_thread_fence

Synopsis

Defined in header <atomic>
extern "C" void atomic_thread_fence( std::memory_order order ) noexcept; (since C++11)

Establishes memory_synchronization_ordering of non-atomic and relaxed atomic accesses, as instructed by order, without an associated atomic operation.

Fence-atomic synchronization

A release fence F in thread A synchronizes-with atomic acquire_operation Y in thread B, if

* there exists an atomic store X (with any memory order)
* Y reads the value written by X (or the value would be written by release_sequence_headed_by_X if X were a release operation)
* F is sequenced-before X in thread A

In this case, all non-atomic and relaxed atomic stores that are sequenced-before F in thread A will happen-before all non-atomic and relaxed atomic loads from the same locations made in thread B after Y.

Atomic-fence synchronization

An atomic release_operation X in thread A synchronizes-with an acquire fence F in thread B, if

* there exists an atomic read Y (with any memory order)
* Y reads the value written by X (or by the release_sequence_headed_by_X)
* Y is sequenced-before F in thread B

In this case, all non-atomic and relaxed atomic stores that are sequenced-before X in thread A will happen-before all non-atomic and relaxed atomic loads from the same locations made in thread B after F.

Fence-fence synchronization

A release fence FA in thread A synchronizes-with an acquire fence FB in thread B, if

* There exists an atomic object M,
* There exists an atomic write X (with any memory order) that modifies M in thread A
* FA is sequenced-before X in thread A
* There exists an atomic read Y (with any memory order) in thread B
* Y reads the value written by X (or the value would be written by release_sequence_headed_by_X if X were a release operation)
* Y is sequenced-before FB in thread B

In this case, all non-atomic and relaxed atomic stores that are sequenced-before FA in thread A will happen-before all non-atomic and relaxed atomic loads from the same locations made in thread B after FB

Parameters

order - the memory ordering executed by this fence

Return value

(none)

Notes

atomic_thread_fence imposes stronger synchronization constraints than an atomic store operation with the same std::memory_order. While an atomic store-release operation prevents all preceding writes from moving past the store-release, an atomic_thread_fence with memory_order_release ordering prevents all preceding writes from moving past all subsequent stores.
Fence-fence synchronization can be used to add synchronization to a sequence of several relaxed atomic operations, for example

  //Global
  std::string computation(int);
  void print( std::string );

  std::atomic<int> arr[3] = { -1, -1, -1 };
  std::string data[1000] //non-atomic data

  // Thread A, compute 3 values
  void ThreadA( int v0, int v1, int v2 )
  {
  //assert( 0 <= v0, v1, v2 < 1000 );
  data[v0] = computation(v0);
  data[v1] = computation(v1);
  data[v2] = computation(v2);
  std::atomic_thread_fence(std::memory_order_release);
  std::atomic_store_explicit(&arr[0], v0, std::memory_order_relaxed);
  std::atomic_store_explicit(&arr[1], v1, std::memory_order_relaxed);
  std::atomic_store_explicit(&arr[2], v2, std::memory_order_relaxed);
  }

  // Thread B, prints between 0 and 3 values already computed.
  void ThreadB()
  {
  int v0 = std::atomic_load_explicit(&arr[0], std::memory_order_relaxed);
  int v1 = std::atomic_load_explicit(&arr[1], std::memory_order_relaxed);
  int v2 = std::atomic_load_explicit(&arr[2], std::memory_order_relaxed);
  std::atomic_thread_fence(std::memory_order_acquire);
  // v0, v1, v2 might turn out to be -1, some or all of them.
  // otherwise it is safe to read the non-atomic data because of the fences:
  if( v0 != -1 ) { print( data[v0] ); }
  if( v1 != -1 ) { print( data[v1] ); }
  if( v2 != -1 ) { print( data[v2] ); }
  }

Examples

Scan an array of mailboxes, and process only the ones intended for us, without unnecessary synchronization. This example uses atomic-fence synchronization.
// Run this code

  const int num_mailboxes = 32;
  std::atomic<int> mailbox_receiver[num_mailboxes];
  std::string mailbox_data[num_mailboxes];

  // The writer threads update non-atomic shared data
  // and then update mailbox_receiver[i] as follows
  mailbox_data[i] = ...;
  std::atomic_store_explicit(&mailbox_receiver[i], receiver_id, std::memory_order_release);

  // Reader thread needs to check all mailbox[i], but only needs to sync with one
  for (int i = 0; i < num_mailboxes; ++i) {
      if (std::atomic_load_explicit(&mailbox_receiver[i], std::memory_order_relaxed) == my_id) {
          std::atomic_thread_fence(std::memory_order_acquire); // synchronize with just one writer
          do_work( mailbox_data[i] ); // guaranteed to observe everything done in the writer thread before
                      // the atomic_store_explicit()
      }
   }