| 322 |
|
* for appends, so can only be removed later when other nodes are |
| 323 |
|
* appended. (2) We cannot necessarily unlink s given a |
| 324 |
|
* predecessor node that is matched (including the case of being |
| 325 |
< |
* cancelled): the predecessor may already be already unspliced, |
| 326 |
< |
* in which case some previous reachable node may still point to |
| 327 |
< |
* s. (For further explanation see Herlihy & Shavit "The Art of |
| 325 |
> |
* cancelled): the predecessor may already be unspliced, in which |
| 326 |
> |
* case some previous reachable node may still point to s. |
| 327 |
> |
* (For further explanation see Herlihy & Shavit "The Art of |
| 328 |
|
* Multiprocessor Programming" chapter 9). Although, in both |
| 329 |
|
* cases, we can rule out the need for further action if either s |
| 330 |
|
* or its predecessor are (or can be made to be) at, or fall off |
| 357 |
|
* Because the sweepVotes estimate is conservative, and because |
| 358 |
|
* nodes become unlinked "naturally" as they fall off the head of |
| 359 |
|
* the queue, and because we allow votes to accumulate even while |
| 360 |
< |
* sweeps are in progress, there are typically signficantly fewer |
| 360 |
> |
* sweeps are in progress, there are typically significantly fewer |
| 361 |
|
* such nodes than estimated. Choice of a threshold value |
| 362 |
|
* balances the likelihood of wasted effort and contention, versus |
| 363 |
|
* providing a worst-case bound on retention of interior nodes in |
