255 |
|
* to the entry points or loop headers of a few methods. A |
256 |
|
* potential disadvantage is that these few remaining fences are |
257 |
|
* not easily optimized away by compilers under exclusively |
258 |
< |
* single-thread use. It requires some care avoid volatile mode |
259 |
< |
* reads of other fields. (Note that the memory semantics of a |
260 |
< |
* reference dependently read in plain mode exactly once are |
258 |
> |
* single-thread use. It requires some care to avoid volatile |
259 |
> |
* mode reads of other fields. (Note that the memory semantics of |
260 |
> |
* a reference dependently read in plain mode exactly once are |
261 |
|
* equivalent to those for atomic opaque mode.) Iterators and |
262 |
|
* other traversals encounter each node and value exactly once. |
263 |
|
* Other operations locate an element (or position to insert an |