return &Arena{a: runtime_arena_newArena()}
}

// Free frees the arena (and all objects allocated from the arena) so that
// memory backing the arena can be reused fairly quickly without garbage
// collection overhead. Applications must not call any method on this
// arena after it has been freed.
func (a *Arena) Free() {
	runtime_arena_arena_Free(a.a)
	a.a = nil
}

   *
   * <p>Although all tasks are immediately executed in the thread that submitted the task, this
   * {@code ExecutorService} imposes a small locking overhead on each task submission in order to
   * implement shutdown and termination behavior.
   *
   * <p>Because of the nature of single-thread execution, the methods {@code scheduleAtFixedRate}

   *
   * <p>Although all tasks are immediately executed in the thread that submitted the task, this
   * {@code ExecutorService} imposes a small locking overhead on each task submission in order to
   * implement shutdown and termination behavior.
   *
   * <p>Because of the nature of single-thread execution, the methods {@code scheduleAtFixedRate}

  }

  /**
   * Returns the current elapsed time shown on this stopwatch, expressed in the desired time unit,
   * with any fraction rounded down.
   *
   * <p><b>Note:</b> the overhead of measurement can be more than a microsecond, so it is generally
   * not useful to specify {@link TimeUnit#NANOSECONDS} precision here.
   *

  }

  /**
   * Returns the current elapsed time shown on this stopwatch, expressed in the desired time unit,
   * with any fraction rounded down.
   *
   * <p>Note that the overhead of measurement can be more than a microsecond, so it is generally not
   * useful to specify {@link TimeUnit#NANOSECONDS} precision here.
   *
   * @since 14.0 (since 10.0 as {@code elapsedTime()})
   */

The *Secure Channel* splits the object content into chunks of a fixed size of `65536` bytes. The last chunk may be smaller to avoid adding additional overhead and is treated specially to prevent truncation attacks. The nonce value is 96 bits long and generated randomly per object / multi-part part. The *Secure Channel* supports plaintexts up to `65536 * 2^32 = 256 TiB`.

#### Randomness

   * been shutdown).
   *
   * <p>Although all tasks are immediately executed in the thread that submitted the task, this
   * {@code ExecutorService} imposes a small locking overhead on each task submission in order to
   * implement shutdown and termination behavior.
   *
   * <p>The implementation deviates from the {@code ExecutorService} specification with regards to

### Multi destination replication

The replication design for multiple sites works in a similar manner as described above for two site scenario. However there are some
important exceptions.

                }
                // == 5*128 = 640

                // . and .. entries = 200 byte (includes alignment)

                // + 64 byte header
                // + 8 byte query response overhead
                // + 110 bytes entry
                // -> 1022 predicted message size <= 1023 maximum buffer size
                // 112 bytes to alignment
                // -> aligned to 1024 > 1023 maximum buffer size

// in tests only.)
TF_CAPI_EXPORT int TF_PickUnusedPortOrDie(void);

// Fast path method that makes constructing a single scalar tensor require less
// overhead and copies.
TF_CAPI_EXPORT extern TFE_TensorHandle* TFE_NewTensorHandleFromScalar(
    TF_DataType data_type, void* data, size_t len, TF_Status* status);

// Specify the server_def that enables collective ops.